1 Introduction
2 Abbreviations
3 Study objectives
4 Changes from the Statistical Analysis Plan
5 Set-up
6 Descriptive analyses
- 6.1 Data overview
7 Baseline population have a low cardiovascular risk profile.
- 7.0.1 Modifiable CVD risk factors and inflammatory and viral markers over time
8 Risk of asymptomatic CVD: surrogate markers CIMT and PWV
9 Summary of clinical and sub-clinical events
10 Incidence rate of clinical and subclinical events are ~2-folds higher in PLWH.
11 Incidence rate of systemic events (mostly hypertension and hypercholesterolaemia) was prevalent in both groups.
12 Mediation analysis
13 Additional things for discussion

1 Introduction

This document largely follows the structure / manuscript plan set our in Laura’s email from 30 November 2022.

Where the narrative set out in that email conflicted with the results, this report was amended accordingly.

2 Abbreviations

CBD - cerebrovascular disease

CRF - case report form

CVD - cardiovascular disease

GLM - Generalised Linear Model; basically the standard regression models, including linear, logistic and Poisson regression models as special cases.

LMM - Linear Mixed Models; a generalisation of linear regression models to longitdunal / panel data; repated measurements for each individual are accommodated throught he specification of a correlation structure and random effects for subject-specific effects.

LTFU - Lost to follow-up or Loss to follow-up.

PLWH - People living with HIV, ART patient participants in RHICCA. Controls, or healthy controls, refer to HIV free study participants.

RHICCA - Reactivation of herpesviruses and inflammation as cardiovascular and cerebrovascular risk factors in antiretroviral initiators, in an African HIV population

3 Study objectives

The RHICCA study will test the hypothesis that immune activation, and in turn dysfunction, driven by HIV and reactivation of latent herpesvirus infections leads to increased CBD/CVD risk in adults aged > 35 years in SSA. We will address this through the following objectives;

To determine if vasculopathy progression or occurrence of new-onset clinically significant vasculopathy is higher in adults aged > 35 years with HIV infection on ART compared to those without HIV
To determine if vasculopathy progression or occurrence of new-onset clinically significant vasculopathy is higher in adults aged > 35 years with viral antigenaemia or chronic immune activation compared to those without viral antigenaemia or chronic immune activation. Specifically, we will determine if vasculopathy is higher:
1. in PLWH with reactivated latent herpes viral infection, compared to those without reactivated latent herpes viral infection.
2. in PLWH in the top 20% of markers for immune activation, immune senescence, inflammation or coagulation compared to the bottom 20%
3. in PLWH with incomplete virological suppression or virological resurgence of HIV, compared to those with suppressed HIV plasma viral load

The secondary study objectives are to determine if viral antigenaemia or chronic immune activation increase occurrence of the following clinical events: 1) stroke, 2) myocardial infarction (MI), 3) angina (excluding MI), 4) peripheral vascular disease (PVD), 5) all-cause death/vascular-related death and 6) immune reconstitution vasculopathy.

4 Changes from the Statistical Analysis Plan

RHICCA is an observational study and much of the analysis is meant to be exploratory. As such, analyses differed in several ways from the SAP that was drafted. Major differences include:

For objective 1, linear mixed models only are used to analyse associations with PWV.
For objective 2b, the natural logarithm of ICAM-1, CD163 and IL-6 were used as predictors rather than categorising according to marker quintiles, which would have been statistically inefficient.
Objective 2c was dropped as virtually all participants were virally suppressed.

5 Set-up

This section contains no output and serves as a container for several helper analysis functions as well as code to read-in and pre-process the data.

# this chunk has code for functions needed for downstream analyses

framinghamScore<-function(dat){
  # Based on equations in Appendix of D'Agostino et alk (2008) 10.1161/CIRCULATIONAHA.107.699579
  # Input 'dat' is a data frame with same column names as rhiccaDatWide in code much further below.
  # The column names will be renamed:
  #  crf.gender --> sex: "Male" or "Female"
  #  crf.age --> age: in years
  #  lims.HDL_Cholesterol_enrolment --> cholHDL = HDL cholesterol: in mmol/L (will be converted to mg/dL for model)
  #  lims.Cholesterol_enrolment --> cholTot = Total cholesterol: in mmol/L (will be converted to mg/dL for model)
  #  crf.bp_treat --> bp_treat = systolic blood pressure treatment: yes/no
  #  physio.bp_sys_enrolment --> bp_sys = systolic blood pressure: in mmHg
  #  crf.smoke --> smoker = smoking status: yes/no
  #  lims.Glucose_enrolment --> diabetes = diabetes status: yes/no (using RHICCA diabetes cut-off of 11.1 mmol/L)
  
  dat<-dat %>%
    dplyr::mutate(
      sex=crf.gender,
      age=crf.age,
      cholHDL=lims.HDL_Cholesterol_enrolment*38.67,
      cholTot=lims.Cholesterol_enrolment*38.67,
      bp_treat=ifelse(crf.bp_treat==1,"yes","no"),
      bp_sys=physio.bp_sys_enrolment,
      smoker=crf.smoke,
      diabetes=ifelse(lims.Glucose_enrolment>=11.1,"yes","no")
    ) %>%
    dplyr::mutate(
      framHamScore=1-case_when(
        sex=="Female" & bp_treat=="no"~0.95012^exp(2.32888*log(age)+1.20904*log(cholTot)-0.70833*log(cholHDL)+2.76157*log(bp_sys)+0.52873*ifelse(smoker=="yes",1,0)+0.69154*ifelse(diabetes=="yes",1,0)-26.1931),
        sex=="Female" & bp_treat=="yes"~0.95012^exp(0.95012^exp(2.32888*log(age)+1.20904*log(cholTot)-0.70833*log(cholHDL)+2.82263*log(bp_sys)+0.52873*ifelse(smoker=="yes",1,0)+0.69154*ifelse(diabetes=="yes",1,0)-26.1931)-26.1931),
        sex=="Male" & bp_treat=="no"~0.88936^exp(3.06117*log(age)+1.12370*log(cholTot)-0.93263*log(cholHDL)+1.93303*log(bp_sys)+0.65451*ifelse(smoker=="yes",1,0)+0.57367*ifelse(diabetes=="yes",1,0)-23.9802),
        sex=="Male" & bp_treat=="yes"~0.88936^exp(3.06117*log(age)+1.12370*log(cholTot)-0.93263*log(cholHDL)+1.99881*log(bp_sys)+0.65451*ifelse(smoker=="yes",1,0)+0.57367*ifelse(diabetes=="yes",1,0)-23.9802)
      )
    )
  
  return(dat$framHamScore)
}

6 Descriptive analyses

# create a big, long data frame combining constant, physiological, PWV, cIMT, ABPI, LIMS data
allPids<-unique(c(constDat$pid,
                  physioDatLong$pid,
                  pwvDatLong$pid,
                  cimtDatLong$pid,
                  abpiDatLong$pid,
                  limsDatLong$pid,
                  lumDatLong$pid))
allVisits<-c("enrolment",paste(sep="",seq(3,36,by=3),"months"),"unscheduled")

gr<-expand.grid(allPids,allVisits)

rhiccaDatLong<-data.frame(
  pid=gr[,1],
  visit=gr[,2]
)

rhiccaDatLong<-rhiccaDatLong %>%
  mutate(art=ifelse(visit=="enrolment",0,1))

addColPrefix<-function(dat,prefix,exclude=character(0),glue="."){
  colnames(dat)[!is.element(colnames(dat),set=exclude)]<-paste(sep=glue,prefix,colnames(dat)[!is.element(colnames(dat),set=exclude)])
  return(dat)
}

constDatTmp<-constDat %>%
  addColPrefix(prefix="crf",exclude=c("pid"))

cardioDatTmp<-cardioDat %>%
  addColPrefix(prefix="cardio",exclude=c("pid"))

physioDatLongTmp <- physioDatLong %>%
  dplyr::select(!c(ptype,physioTmpID)) %>%
  addColPrefix(prefix="physio",exclude=c("pid","timePoint"))

pwvDatLongTmp <- pwvDatLong %>%
  dplyr::select(!c(vtype,test_time,src,pwvTmpID,instanceid,project_dsid)) %>%
  addColPrefix(prefix="pwv",exclude=c("pid","timePoint"))

cimtDatLongTmp <- cimtDatLong %>%
  dplyr::select(!c(vtype,src,cimtTmpID,instanceid,project_dsid)) %>%
  addColPrefix(prefix="cimt",exclude=c("pid","timePoint"))

abpiDatLongTmp <- abpiDatLong %>%
  dplyr::select(!c(vtype,src,abpiTmpID,instanceid,project_dsid)) %>%
  addColPrefix(prefix="abpi",exclude=c("pid","timePoint"))

limsDatLongTmp <- limsDatLong %>%
  dplyr::select(!pid.x) %>%
  addColPrefix(prefix="lims",exclude=c("pid.y","timePoint"))

lumDatLongTmp <- lumDatLong %>%
  dplyr::select(!c(SampleID,timepoint)) %>%
  addColPrefix(prefix="lum",exclude=c("pid","timePoint"))

rhiccaDatLong<- rhiccaDatLong %>%
  left_join(y=constDatTmp,by="pid") %>%
  left_join(y=physioDatLongTmp,by=c("pid","visit"="timePoint")) %>%
  left_join(y=pwvDatLongTmp,by=c("pid","visit"="timePoint")) %>%
  left_join(y=cimtDatLongTmp,by=c("pid","visit"="timePoint")) %>%
  left_join(y=abpiDatLongTmp,by=c("pid","visit"="timePoint")) %>%
  left_join(y=limsDatLongTmp,by=c("pid"="pid.y","visit"="timePoint")) %>%
  left_join(y=lumDatLongTmp,by=c("pid","visit"="timePoint"))

rhiccaDatLong <- rhiccaDatLong %>%
  mutate(physio.bmi=physio.weight/(crf.height/100)^2,
         physio.waisthip=physio.waist/physio.hip,
         visitNum=ifelse(visit=="enrolment",0,as.integer(gsub(pattern="months",replacement="",visit))),
         visitNumExactPwv=(dmy(pwv.data_date)-dmy(crf.data_date))/dmonths(1),
         visitNumExactCimt=(dmy(cimt.data_date)-dmy(crf.data_date))/dmonths(1),
         visitNumExactAbpi=(dmy(abpi.data_date)-dmy(crf.data_date))/dmonths(1),
         visitNumExactPhysio=(dmy(physio.data_date)-dmy(crf.data_date))/dmonths(1)
  ) %>%
  mutate(physio.bmi_cat=case_when(physio.bmi<18.5~"underweight",
                                  physio.bmi>=18.5 & physio.bmi<25~"normal",
                                  physio.bmi>=25 & physio.bmi<30~"overweight",
                                  physio.bmi>=30~"obese"))

naCount<-rowSums(is.na(rhiccaDatLong))
idxRm<-which(naCount==ncol(rhiccaDatLong)-ncol(constDat))
if(length(idxRm)>0){rhiccaDatLong<-rhiccaDatLong[-idxRm,]}

egfrCalc<-function(creat_mumolPerL,sex,age,eth="black"){
  n<-length(creat_mumolPerL)
  if(n!=length(sex) | n!=length(age)){stop("creatMmolL, sex, age need to be vectors of the same length!")}
  
  # using thr 2021 CK-EPI eGFRcr(AS) equation from https://doi.org/110.1056/NEJMoa2102953
  # https://www.kidney.org/content/ckd-epi-creatinine-equation-2021
  creat_mgPerdl<-creat_mumolPerL*0.0113/ifelse(sex=="Female",0.7,0.9)
  res<-142 * (apply(X=cbind(creat_mgPerdl,rep(1,n)),FUN=min,MARGIN=1)^ifelse(sex=="Female",-0.241,-0.302)) * (apply(X=cbind(creat_mgPerdl,rep(1,n)),FUN=max,MARGIN=1)^(-1.200)) * (0.9938^age) * ifelse(sex=="Female",1.012,1) 
  
  # using the CKD-EPI equation from https://doi.org/10.7326/0003-4819-150-9-200905050-00006
  # const<-case_when(eth=="black" & sex=="Female"~166,eth=="black" & sex=="Male"~163,eth=="white" & sex=="Female"~144,eth=="white" & sex=="Male"~141)
  # 
  # creat<-creat_mumolPerL*0.0113
  # # creat<-creat_mumolPerL/11.312222
  # creat<-creat/ifelse(sex=="Female",0.7,0.9)
  # 
  # expo<-case_when(sex=="Female" & creat<=62~(-0.329),sex=="Female" & creat>62~(-1.209),sex=="Male" & creat<=80~(-0.411),sex=="Male" & creat>80~(-1.209))
  # 
  # res<-const*creat^expo*0.993^age
  # #res<-141 * (apply(X=cbind(creat,rep(1,n)),FUN=min,MARGIN=1)^ifelse(sex=="Female",-0.329,-0.411)) * (apply(X=cbind(creat,rep(1,n)),FUN=max,MARGIN=1)^(-1.209)) * (0.993^age) * ifelse(sex=="Female",1.018,1) * 1
  # # # # assuming everyone is black
  
  # using the 4-variable MDRD equations from https://doi.org/10.7326/0003-4819-145-4-200608150-00004
  #creat_mgPerdl<-reat_mumolPerL*0.0113
  #res<-175 * creat_mgPerdl^(-0.203) * age^(-0.203) * ifelse(eth=="black",1.212,1) * ifelse(sex=="Female",0.742,1) 
  return(res)
}

rhiccaDatLong <- rhiccaDatLong %>%
  mutate(eGFR=egfrCalc(creat_mumolPerL=lims.Creatinine,sex=crf.gender,age=crf.age))

rm(gr,naCount,idxRm,allPids,allVisits,constDatTmp,physioDatLongTmp,pwvDatLongTmp,cimtDatLongTmp,abpiDatLongTmp,limsDatLongTmp,lumDatLongTmp)

rhiccaDatLong<-rhiccaDatLong %>%
  filter(!(pid %in% pidsToRemove))

rhiccaDatWide <- rhiccaDatLong %>%
  filter(visit!="unscheduled") %>%
  pivot_wider(names_from=visit,values_from=!c(pid,starts_with("crf."),starts_with("cardio.")))

rhiccaDatWide <- rhiccaDatWide %>%
  mutate(
    iris=factor(case_when(
      pwv.pwv_enrolment>12 ~ "pre-existing vasculopathy",
      pwv.pwv_enrolment<=12 & pwv.pwv_6months<=12 ~ "no vasculopathy",
      pwv.pwv_enrolment<=12 & pwv.pwv_6months>12 ~ "new onset vasculopathy"
    ),levels=c("no vasculopathy","pre-existing vasculopathy","new onset vasculopathy")),
    ltfu_enrolment=ifelse(is.na(physio.data_date_enrolment) & is.na(physio.bmi_enrolment) & is.na(physio.temp_enrolment) & is.na(physio.hr_enrolment),1,0),
    ltfu_6months=ifelse(is.na(physio.data_date_6months) & is.na(physio.bmi_6months) & is.na(physio.temp_6months) & is.na(physio.hr_6months),1,0),
    ltfu_12months=ifelse(is.na(physio.data_date_12months) & is.na(physio.bmi_12months) & is.na(physio.temp_12months) & is.na(physio.hr_12months),1,0),
    ltfu_18months=ifelse(is.na(physio.data_date_18months) & is.na(physio.bmi_18months) & is.na(physio.temp_18months) & is.na(physio.hr_18months),1,0),
    ltfu_24months=ifelse(is.na(physio.data_date_24months) & is.na(physio.bmi_24months) & is.na(physio.temp_24months) & is.na(physio.hr_24months),1,0),
    ltfu_30months=ifelse(is.na(physio.data_date_30months) & is.na(physio.bmi_30months) & is.na(physio.temp_30months) & is.na(physio.hr_30months),1,0),
    ltfu_24months_reason=ltfuDat$reason[match(rhiccaDatWide$pid,ltfuDat$pid)],
    ltfu_24months_ineligDetails=ltfuDat$ineligibility_details[match(rhiccaDatWide$pid,ltfuDat$pid)]
  )

rhiccaDatLong<-rhiccaDatLong %>%
  add_column(iris=rhiccaDatWide$iris[match(rhiccaDatLong$pid,rhiccaDatWide$pid)],
             HIV_VL_base=rhiccaDatWide$lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment[match(rhiccaDatLong$pid,rhiccaDatWide$pid)])

idsNoBase<-rhiccaDatLong %>%
  filter(is.na(crf.ptype)) %>%
  dplyr::select(pid) %>%
  unique()
idsNoBase<-setdiff(idsNoBase$pid,idsNoVL$pid)
write.csv(idsNoBase,file=paste(sep="","../output/",gsub(pattern="-",replacement="",Sys.Date()),"/rhiccaDat_noBaselineCrfData",gsub(pattern="-",replacement="",Sys.Date()),".csv"),row.names=F)

idxBMIunder<-which(is.element(rhiccaDatLong$pid,set=rhiccaDatWide$pid[rhiccaDatWide$physio.bmi_cat_enrolment=="underweight"]))
idxBMInormal<-which(is.element(rhiccaDatLong$pid,set=rhiccaDatWide$pid[rhiccaDatWide$physio.bmi_cat_enrolment=="normal"]))
idxBMIover<-which(is.element(rhiccaDatLong$pid,set=rhiccaDatWide$pid[rhiccaDatWide$physio.bmi_cat_enrolment=="overweight"]))
idxBMIobese<-which(is.element(rhiccaDatLong$pid,set=rhiccaDatWide$pid[rhiccaDatWide$physio.bmi_cat_enrolment=="obese"]))

idxHypertensionNormal<-which(is.element(rhiccaDatLong$pid,set=rhiccaDatWide$pid[rhiccaDatWide$physio.bp_group_enrolment=="Normotensive"]))
idxHypertensionPre<-which(is.element(rhiccaDatLong$pid,set=rhiccaDatWide$pid[rhiccaDatWide$physio.bp_group_enrolment=="Pre-hypertensive"]))
idxHypertensionHyper<-which(is.element(rhiccaDatLong$pid,set=rhiccaDatWide$pid[rhiccaDatWide$physio.bp_group_enrolment=="Hypertensive"]))

idxPriorART<-which(!is.na(rhiccaDatWide$crf.artnaive) & rhiccaDatWide$crf.artnaive=="prior ART" & rhiccaDatWide$crf.ptype=="Patient")
idxPriorARTNoVL<-which(!is.na(rhiccaDatWide$crf.artnaive) & rhiccaDatWide$crf.artnaive=="prior ART" & rhiccaDatWide$crf.ptype=="Patient" & !is.na(rhiccaDatWide$lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment) & rhiccaDatWide$lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment==150)
pidsPriorART<-rhiccaDatWide$pid[idxPriorART]
pidsPriorARTNoVL<-rhiccaDatWide$pid[idxPriorARTNoVL]

art<-data.frame(pid=rhiccaDatWide$pid,art=paste(sep=";",rhiccaDatWide$crf.artRegimen,rhiccaDatWide$crf.artRegimenNew))
art<-art[rhiccaDatWide$crf.ptype=="Patient",]
for(i in 1:nrow(art)){
  tmp<-unlist(strsplit(split=";",art$art[i]))
  tmp<-tmp[!grepl(tmp,pattern="NA")]
  tmp<-sort(unique(tmp))
  idx5A<-which(tmp=="5A")
  if(length(idx5A)==1){tmp<-c(tmp[idx5A],tmp[-idx5A])}
  tmp<-paste(collapse=", ",tmp)
  art$art[i]<-tmp
}
art<-art %>%
  mutate(
    artSimple=case_when(
    is.element(el=art,set=c("5A","5A, 13A"))~art,
    TRUE~"Other"),
    artSimple2Level=factor(case_when(
      grepl(pattern="13A",art)~"ART: Change to dolutegravir",
      TRUE~"ART: No change to dolutegravir"),levels=c("ART: Change to dolutegravir","ART: No change to dolutegravir"))
    )
    
tabArt<-as.data.frame(table(art$art))
tabArt<-tabArt[order(decreasing = T,tabArt$Freq),]

rhiccaDatWide$crf.artRegimen.full<-art$art[match(rhiccaDatWide$pid,art$pid)]
rhiccaDatWide$crf.artRegimen.simple<-art$artSimple[match(rhiccaDatWide$pid,art$pid)]
rhiccaDatWide$crf.artRegimen.2levels<-art$artSimple2Level[match(rhiccaDatWide$pid,art$pid)]

rhiccaDatWide$eGFR_existingCKD<-ifelse(!is.na(rhiccaDatWide$eGFR_enrolment) & rhiccaDatWide$eGFR_enrolment<60,"yes","no")

6.1 Data overview

6.1.1 Data completeness & loss-to-follow-up

After merging the different data tables, there are quite a few missing data points. Table 6.1 summarises the situation.

missSumFun<-function(dat){
  varVect<-c("pwv.pwv","cimt.cimt","abpi.abpi","lims.Cholesterol","lims.Glucose","eGFR","lum.CD163","lum.ICAM1","lum.IL6","lum.ILb")
  
  df<-data.frame(
    nTotal=nrow(dat),
    nBaseline=c(sum(dat$ltfu_enrolment==0),colSums(!is.na(dat[,c(paste(sep="_",varVect,"enrolment"))]))),
    n6months=c(sum(dat$ltfu_6months==0),colSums(!is.na(dat[,c(paste(sep="_",varVect,"6months"))]))),
    n12months=c(sum(dat$ltfu_12months==0),colSums(!is.na(dat[,c(paste(sep="_",varVect,"12months"))]))),
    n18months=c(sum(dat$ltfu_18months==0),colSums(!is.na(dat[,c(paste(sep="_",varVect,"18months"))]))),
    n24months=c(sum(dat$ltfu_24months==0),colSums(!is.na(dat[,c(paste(sep="_",varVect,"24months"))])))
  )
  
  varVect2<-c("lims.HIV_Abbott_Viral_Load_0pt2ml_P","lims.CD4_Absolute")
  dat<-dat[!is.na(dat$crf.ptype) & dat$crf.ptype=="Patient",]
  df2<-data.frame(
    nTotal=nrow(dat),
    nBaseline=c(sum(dat$ltfu_enrolment==0),colSums(!is.na(dat[,c(paste(sep="_",varVect2,"enrolment"))]))),
    n6months=c(sum(dat$ltfu_6months==0),colSums(!is.na(dat[,c(paste(sep="_",varVect2,"6months"))]))),
    n12months=c(sum(dat$ltfu_12months==0),colSums(!is.na(dat[,c(paste(sep="_",varVect2,"12months"))]))),
    n18months=c(sum(dat$ltfu_18months==0),colSums(!is.na(dat[,c(paste(sep="_",varVect2,"18months"))]))),
    n24months=c(sum(dat$ltfu_24months==0),colSums(!is.na(dat[,c(paste(sep="_",varVect2,"24months"))])))
  )
  
  df<-rbind(df,df2)
  
  varNames<-c("Physiological data","PWV","CIMT","ABPI","Cholesterol","Glucose","eGFR","CD163","ICAM-1","IL-6","IL-1b","Physiological data (PLWH only)","HIV VL (PLWH only)","CD4+ (PLWH only)")
  rownames(df)<-varNames
  return(df)
}

datMiss<-missSumFun(rhiccaDatWide)
datMissNoMissCrf<-missSumFun(rhiccaDatWide %>% filter(!is.na(crf.ptype)))

datMiss %>%
  kable(caption="Numbers of non-missing data at the various visits for CRF, PWV, CIMT and ABPI.",col.names = c("Total","Baseline","6 month","12 months","18 months","24 months")) %>%
  kable_styling(full_width = F)

Table 6.1: Numbers of non-missing data at the various visits for CRF, PWV, CIMT and ABPI.
	Total	Baseline	6 month	12 months	18 months	24 months
Physiological data	923	903	791	765	726	676
PWV	923	904	779	754	720	675
CIMT	923	915	0	0	0	671
ABPI	923	911	784	759	726	682
Cholesterol	923	908	0	0	0	641
Glucose	923	909	0	0	0	679
eGFR	923	898	0	0	0	683
CD163	923	744	746	0	0	0
ICAM-1	923	746	745	0	0	0
IL-6	923	746	746	0	0	0
IL-1b	923	745	746	0	0	0
Physiological data (PLWH only)	713	713	611	584	559	520
HIV VL (PLWH only)	713	692	600	565	0	0
CD4+ (PLWH only)	713	703	603	579	0	528

missDatList<-list()
missDatList$pwv<-list()
missDatList$cimt<-list()
missDatList$abpi<-list()
missDatList$cholesterol<-list()
missDatList$triglycerides<-list()
missDatList$glucose<-list()
missDatList$VL<-list()

missDatList$pwv$baseline<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$pwv.pwv_enrolment)]
missDatList$pwv$fu24m<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$pwv.pwv_24months)]

missDatList$cimt$baseline<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$cimt.cimt_enrolment)]
missDatList$cimt$fu24m<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$cimt.cimt_24months)]

missDatList$abpi$baseline<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$abpi.abpi_enrolment)]
missDatList$abpi$fu24m<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$abpi.abpi_24months)]

missDatList$cholesterol$baseline<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$lims.Cholesterol_enrolment)]
missDatList$cholesterol$fu24m<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$lims.Cholesterol_24months)]

missDatList$triglycerides$baseline<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$lims.Triglycerides_enrolment)]
missDatList$triglycerides$fu24m<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$lims.Triglycerides_24months)]

missDatList$glucose$baseline<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$lims.Glucose_enrolment)]
missDatList$glucose$fu24m<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$lims.Glucose_enrolment)]

missDatList$VL$baseline<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)]
missDatList$VL$fu24m<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$crf.ptype) & is.na(rhiccaDatWide$lims.HIV_Abbott_Viral_Load_0pt2ml_P_24months)]

save(missDatList,file=paste(sep="","../output/",gsub(pattern="-",replacement="",Sys.Date()),"/rhiccaDat_allMerged_missingBaseline24monthData",gsub(pattern="-",replacement="",Sys.Date()),".RData"))

#####################
## clinical events ##
#####################

# confirm these!
clinEventsDat$PID[clinEventsDat$PID=="RE10930"]<-"RE1093O"
clinEventsDat$PID[clinEventsDat$PID=="RE12290"]<-"RE1229O"
#clinEventsDat$PID[clinEventsDat$PID=="RE1628A"]<-"RE1628V"
clinEventsDat$PID[clinEventsDat$PID=="RE14210"]<-"RE1421O"
clinEventsDat$PID[clinEventsDat$PID=="RE1118L"]<-"RE11181"
clinEventsDat$PID[clinEventsDat$PID=="RE1230I"]<-"RE12301"
clinEventsDat$PID[clinEventsDat$PID=="RE13281"]<-"RE1328I"
clinEventsDat$PID[clinEventsDat$PID=="RE1815H"]<-"RE1815P"

for(i in 1:nrow(clinEventsDat)){
  if(clinEventsDat$PID[i]==""){clinEventsDat$PID[i]<-clinEventsDat$PID[i-1]} # file is organised in this way that multiple events for the same participants do not repeat the PID but are in subsequent rows....
}

clinEventsDat$gender<-rhiccaDatWide$crf.gender[match(clinEventsDat$PID,rhiccaDatWide$pid)]
clinEventsDat$age_group<-rhiccaDatWide$crf.age_group[match(clinEventsDat$PID,rhiccaDatWide$pid)]
clinEventsDat$ptype<-rhiccaDatWide$crf.ptype[match(clinEventsDat$PID,rhiccaDatWide$pid)]
clinEventsDat$artnaive<-rhiccaDatWide$crf.artnaive[match(clinEventsDat$PID,rhiccaDatWide$pid)]

clinEventsDatFull<-clinEventsDat %>%
  mutate(degreeOfCertainty = gsub(pattern="^ ",replacement="",gsub(pattern=" $",replacement="",tolower(DEGREE.OF.CERTAINITY)))) %>%
  mutate(DIAGNOSIS = gsub(pattern="^ ",replacement="",gsub(pattern=" $",replacement="",tolower(DIAGNOSIS)))) %>%
  # mutate(diagnosisSimplified = case_when(
  #   grepl(DIAGNOSIS,pattern="death") ~ "death",
  #   grepl(DIAGNOSIS,pattern="anaemia") ~ "anaemia",
  #   grepl(DIAGNOSIS,pattern="tb$") ~ "tb",
  #   grepl(DIAGNOSIS,pattern="hypertension") ~ "hypertension",
  #   DIAGNOSIS=="herpes zoster, disseminated" ~ "disseminated herpes zoster",
  #   TRUE ~ DIAGNOSIS
  # )) %>%
  dplyr::select(!DEGREE.OF.CERTAINITY)

clinEventsDat<-clinEventsDatFull %>%
  filter(degreeOfCertainty!="not an event")

clinEventsDat$category<-case_when(
  grepl(clinEventsDat$DIAGNOSIS,pattern="aids-cancer")~"Cancer",
  grepl(clinEventsDat$DIAGNOSIS,pattern="anaemia")~"Systemic disease",
  grepl(clinEventsDat$DIAGNOSIS,pattern="candidiasis")~"Infection",
  grepl(clinEventsDat$DIAGNOSIS,pattern="renal")~"Systemic disease",
  grepl(clinEventsDat$DIAGNOSIS,pattern="hypertension")~"Systemic disease",
  grepl(clinEventsDat$DIAGNOSIS,pattern="liver")~"Systemic disease",
  grepl(clinEventsDat$DIAGNOSIS,pattern="hzv")~"Infection",
  grepl(clinEventsDat$DIAGNOSIS,pattern="tuberculosis")~"Infection",
  grepl(clinEventsDat$DIAGNOSIS,pattern="pneumonia")~"Infection",
  grepl(clinEventsDat$DIAGNOSIS,pattern="pneumocystis")~"Infection",
  grepl(clinEventsDat$DIAGNOSIS,pattern="salmonella")~"Infection",
  clinEventsDat$DIAGNOSIS=="death" & grepl(clinEventsDat$Cause.of.death.1_ERC_summary,pattern="Systemic")~"Systemic disease",
  clinEventsDat$DIAGNOSIS=="death" & grepl(clinEventsDat$Cause.of.death.1_ERC_summary,pattern="Cancer")~"Cancer",
  clinEventsDat$DIAGNOSIS=="death" & grepl(clinEventsDat$Cause.of.death.1_ERC_summary,pattern="Infections")~"Infection",
  clinEventsDat$DIAGNOSIS=="death" & grepl(clinEventsDat$Cause.of.death.1_ERC_summary,pattern="Cardiovascular")~"CVD",
  clinEventsDat$DIAGNOSIS=="death" & grepl(clinEventsDat$Cause.of.death.1_summary,pattern="Systemic")~"Systemic disease",
  clinEventsDat$DIAGNOSIS=="death" & grepl(clinEventsDat$Cause.of.death.1_summary,pattern="Cancer")~"Cancer",
  clinEventsDat$DIAGNOSIS=="death" & grepl(clinEventsDat$Cause.of.death.1_summary,pattern="Infections")~"Infection",
  clinEventsDat$DIAGNOSIS=="death" & grepl(clinEventsDat$Cause.of.death.1_summary,pattern="Cardiovascular")~"CVD",
  TRUE~"Undetermined"
)

#clinEventsDat$Category<-factor(clinEventsDat$Category)

#barPlotFun(dat=clinEventsDat,var="diagnosisSimplified",groupVar="ptype",labels=c("Control","ART patient"),xLab="Clinical events",legPos = "topleft",legDirHoriz = F)

rhiccaDatWide<-rhiccaDatWide %>%
  mutate(
    newOnsetDiabetes=case_when( # data dictionary does not list glucose lowering medication variable??
      lims.Glucose_enrolment>=11.1~"pre-existing diabetes",
      lims.Glucose_enrolment<11.1 & lims.Glucose_24months>=11.1~"new diabetes",
      is.na(lims.Glucose_enrolment) | is.na(lims.Glucose_24months)~NA_character_,
      TRUE~"no diabetes"
    ),
    newOnsetHypercholesterolemia=case_when(
      lims.Cholesterol_enrolment>=5.2~"pre-existing hypercholesterolemia",
      lims.Cholesterol_enrolment<5.2 & lims.Cholesterol_24months>=5.2~"new hypercholesterolemia",
      is.na(lims.Cholesterol_enrolment) | is.na(lims.Cholesterol_24months)~NA_character_,
      TRUE~"no hypercholesterolemia"
    ),
    newOnsetKidneyCondition=case_when(
      eGFR_enrolment<60~"pre-existing kidney condition",
      #eGFR_enrolment<60 & eGFR_6months>=60~"new kidney condition",
      eGFR_enrolment>=60 & eGFR_24months<60~"new kidney condition",
      #is.na(eGFR_enrolment)~NA_character_,
      is.na(eGFR_enrolment) | is.na(eGFR_24months)~NA_character_,
      #is.na(eGFR_6months) & is.na(eGFR_24months)~NA_character_,
      #is.na(eGFR_6months) & (eGFR_enrolment<60 & eGFR_24months<60)~NA_character_,
      #is.na(eGFR_24months) & (eGFR_enrolment<60 & eGFR_6months<60)~NA_character_,
      TRUE~"no kidney condition"
    ),
    newOnsetABPI=case_when(
      abpi.abpi_enrolment<=0.9~"pre-existing ABPI vasculopathy",
      abpi.abpi_enrolment>0.9 & abpi.abpi_24months<=0.9~"new ABPI vasculopathy",
      is.na(abpi.abpi_enrolment) | is.na(abpi.abpi_24months)~NA_character_,
      TRUE~"no ABPI vasculopathy"
    ),
    newOnsetPWV=case_when(
      pwv.pwv_enrolment>12~"pre-existing PWV vasculopathy",
      pwv.pwv_enrolment<=12 & pwv.pwv_24months>12~"new PWV vasculopathy",
      is.na(pwv.pwv_enrolment) | is.na(pwv.pwv_24months)~NA_character_,
      TRUE~"no PWV vasculopathy"
    ),
    newOnsetCIMT=case_when(
      cimt.cimt_enrolment>0.78~"pre-existing CIMT vasculopathy",
      cimt.cimt_enrolment<=0.78 & cimt.cimt_24months>0.78~"new CIMT vasculopathy",
      is.na(cimt.cimt_enrolment) | is.na(cimt.cimt_24months)~NA_character_,
      TRUE~"no PWV vasculopathy"
    ),
    newOnsetCIMT0.9=case_when(
      cimt.cimt_enrolment>0.9~"pre-existing CIMT vasculopathy",
      cimt.cimt_enrolment<=0.9 & cimt.cimt_24months>0.9~"new CIMT vasculopathy",
      is.na(cimt.cimt_enrolment) | is.na(cimt.cimt_24months)~NA_character_,
      TRUE~"no PWV vasculopathy"
    ),
    newOnsetMyocardialInfarction=case_when(
      pid %in% cardioDat$pid[cardioDat$mc_mi==1 & cardioDat$timeSinceBaseline<30]~"pre-existing MI",
      pid %in% cardioDat$pid[cardioDat$mc_mi==1 & cardioDat$timeSinceBaseline>=30]~"new MI",
      pid %in% cardioDat$pid[is.na(cardioDat$timeSinceBaseline)]~NA_character_,
      TRUE~"noMI"
    )
  )

rhiccaDatWide<-rhiccaDatWide %>%
  mutate(
    newOnsetClinEvent=case_when(
      newOnsetDiabetes=="new diabetes"~1,
      newOnsetHypercholesterolemia=="new hypercholesterolemia"~1,
      newOnsetKidneyCondition=="new kidney condition"~1,
      newOnsetABPI=="new ABPI vasculopathy"~1,
      newOnsetPWV=="new PWV vasculopathy"~1,
      newOnsetCIMT=="new CIMT vasculopathy"~1,
      newOnsetMyocardialInfarction=="new MI"~1,
      TRUE~0
    ),
    newOnsetClinEventNoAbpiPwvCimt=case_when(
      newOnsetDiabetes=="new diabetes"~1,
      newOnsetHypercholesterolemia=="new hypercholesterolemia"~1,
      newOnsetKidneyCondition=="new kidney condition"~1,
      newOnsetMyocardialInfarction=="new MI"~1,
      TRUE~0
    )
  )
    
rhiccaDatWide<-rhiccaDatWide %>%
  mutate(
    clinEvent=case_when(
      newOnsetClinEvent==1~1,
      pid %in% clinEventsDat$PID~1,
      TRUE~0
    ),
    clinEventConfirmed=case_when(
      newOnsetClinEventNoAbpiPwvCimt==1~1,
      pid %in% clinEventsDat$PID[clinEventsDat$degreeOfCertainty=="confirmed"]~1,
      TRUE~0
    )
  )

trblShoot<-data.frame(
  clinicalEvent=c("Diabetes","Hypercholesterolemia","Chronic_kidney_disease","Vasculopathy (ABPI)","Vasculopathy (PWV)","Vasculopathy (CIMT)","MI"),
  variable=c("lims.Glucose","lims.Cholesterol","eGFR","ABPI","PWV","CIMT","MI"),
  nTotal=nrow(rhiccaDatWide),
  missingNewOnsetStatus=c(
    sum(is.na(rhiccaDatWide$newOnsetDiabetes)),
    sum(is.na(rhiccaDatWide$newOnsetHypercholesterolemia)),
    sum(is.na(rhiccaDatWide$newOnsetKidneyCondition)),
    sum(is.na(rhiccaDatWide$newOnsetABPI)),
    sum(is.na(rhiccaDatWide$newOnsetPWV)),
    sum(is.na(rhiccaDatWide$newOnsetCIMT)),
    sum(is.na(rhiccaDatWide$newOnsetMyocardialInfarction))
  ),
  missingVarBaseline=c(
    sum(is.na(rhiccaDatWide$lims.Glucose_enrolment)),
    sum(is.na(rhiccaDatWide$lims.Cholesterol_enrolment)),
    sum(is.na(rhiccaDatWide$eGFR_enrolment)),
    sum(is.na(rhiccaDatWide$abpi.abpi_enrolment)),
    sum(is.na(rhiccaDatWide$pwv.pwv_enrolment)),
    sum(is.na(rhiccaDatWide$cimt.cimt_enrolment)),
    NA
  ),
  missingVar6months=c(
    sum(is.na(rhiccaDatWide$lims.Glucose_6months)),
    sum(is.na(rhiccaDatWide$lims.Cholesterol_6months)),
    sum(is.na(rhiccaDatWide$eGFR_6months)),
    sum(is.na(rhiccaDatWide$abpi.abpi_6months)),
    sum(is.na(rhiccaDatWide$pwv.pwv_6months)),
    sum(is.na(rhiccaDatWide$cimt.cimt_6months)),
    NA
  ),
  missingVar24months=c(
    sum(is.na(rhiccaDatWide$lims.Glucose_24months)),
    sum(is.na(rhiccaDatWide$lims.Cholesterol_24months)),
    sum(is.na(rhiccaDatWide$eGFR_24months)),
    sum(is.na(rhiccaDatWide$abpi.abpi_24months)),
    sum(is.na(rhiccaDatWide$pwv.pwv_24months)),
    sum(is.na(rhiccaDatWide$cimt.cimt_24months)),
    NA
  )
)

write.csv(trblShoot,file=paste(sep="","../output/",gsub(pattern="-",replacement="",Sys.Date()),"/rhiccaDat_missingNewOnsetClinicalEvents_",gsub(pattern="-",replacement="",Sys.Date()),".csv"),row.names=F)

rhiccaDatWide<-rhiccaDatWide %>% 
  add_column(clinEventCategory=character(nrow(rhiccaDatWide))) %>%
  add_column(clinEventCategoryConfirmed=character(nrow(rhiccaDatWide))) %>%
  add_column(clinEventCategoryConfirmedSystemicType=character(nrow(rhiccaDatWide)))

rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetDiabetes) & rhiccaDatWide$newOnsetDiabetes=="new diabetes"]<-paste(sep=";",rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetDiabetes) & rhiccaDatWide$newOnsetDiabetes=="new diabetes"],"Systemic disease")
rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetHypercholesterolemia) & rhiccaDatWide$newOnsetHypercholesterolemia=="new hypercholesterolemia"]<-paste(sep=";",rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetHypercholesterolemia) & rhiccaDatWide$newOnsetHypercholesterolemia=="new hypercholesterolemia"],"Systemic disease")
rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetKidneyCondition) & rhiccaDatWide$newOnsetKidneyCondition=="new kidney condition"]<-paste(sep=";",rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetKidneyCondition) & rhiccaDatWide$newOnsetKidneyCondition=="new kidney condition"],"Systemic disease")
rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetABPI) & rhiccaDatWide$newOnsetABPI=="new ABPI vasculopathy"]<-paste(sep=";",rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetABPI) & rhiccaDatWide$newOnsetABPI=="new ABPI vasculopathy"],"CVD")
rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetPWV) & rhiccaDatWide$newOnsetPWV=="new PWV vasculopathy"]<-paste(sep=";",rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetPWV) & rhiccaDatWide$newOnsetPWV=="new PWV vasculopathy"],"CVD")
rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetCIMT) & rhiccaDatWide$newOnsetCIMT=="new CIMT vasculopathy"]<-paste(sep=";",rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetCIMT) & rhiccaDatWide$newOnsetCIMT=="new CIMT vasculopathy"],"CVD")
rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetMyocardialInfarction) & rhiccaDatWide$newOnsetMyocardialInfarction=="new MI"]<-paste(sep=";",rhiccaDatWide$clinEventCategory[!is.na(rhiccaDatWide$newOnsetMyocardialInfarction) & rhiccaDatWide$newOnsetMyocardialInfarction=="new MI"],"CVD")

rhiccaDatWide$clinEventCategoryConfirmed[!is.na(rhiccaDatWide$newOnsetDiabetes) & rhiccaDatWide$newOnsetDiabetes=="new diabetes"]<-paste(sep=";",rhiccaDatWide$clinEventCategoryConfirmed[!is.na(rhiccaDatWide$newOnsetDiabetes) & rhiccaDatWide$newOnsetDiabetes=="new diabetes"],"Systemic disease")
rhiccaDatWide$clinEventCategoryConfirmedSystemicType[!is.na(rhiccaDatWide$newOnsetDiabetes) & rhiccaDatWide$newOnsetDiabetes=="new diabetes"]<-paste(sep=";",rhiccaDatWide$clinEventCategoryConfirmedSystemicType[!is.na(rhiccaDatWide$newOnsetDiabetes) & rhiccaDatWide$newOnsetDiabetes=="new diabetes"],"diabetes")
rhiccaDatWide$clinEventCategoryConfirmed[!is.na(rhiccaDatWide$newOnsetHypercholesterolemia) & rhiccaDatWide$newOnsetHypercholesterolemia=="new hypercholesterolemia"]<-paste(sep=";",rhiccaDatWide$clinEventCategoryConfirmed[!is.na(rhiccaDatWide$newOnsetHypercholesterolemia) & rhiccaDatWide$newOnsetHypercholesterolemia=="new hypercholesterolemia"],"Systemic disease")
rhiccaDatWide$clinEventCategoryConfirmedSystemicType[!is.na(rhiccaDatWide$newOnsetHypercholesterolemia) & rhiccaDatWide$newOnsetHypercholesterolemia=="new hypercholesterolemia"]<-paste(sep=";",rhiccaDatWide$clinEventCategoryConfirmedSystemicType[!is.na(rhiccaDatWide$newOnsetHypercholesterolemia) & rhiccaDatWide$newOnsetHypercholesterolemia=="new hypercholesterolemia"],"hypercholesterolemia")
rhiccaDatWide$clinEventCategoryConfirmed[!is.na(rhiccaDatWide$newOnsetKidneyCondition) & rhiccaDatWide$newOnsetKidneyCondition=="new kidney condition"]<-paste(sep=";",rhiccaDatWide$clinEventCategoryConfirmed[!is.na(rhiccaDatWide$newOnsetKidneyCondition) & rhiccaDatWide$newOnsetKidneyCondition=="new kidney condition"],"Systemic disease")
rhiccaDatWide$clinEventCategoryConfirmedSystemicType[!is.na(rhiccaDatWide$newOnsetKidneyCondition) & rhiccaDatWide$newOnsetKidneyCondition=="new kidney condition"]<-paste(sep=";",rhiccaDatWide$clinEventCategoryConfirmedSystemicType[!is.na(rhiccaDatWide$newOnsetKidneyCondition) & rhiccaDatWide$newOnsetKidneyCondition=="new kidney condition"],"kidney disease")
rhiccaDatWide$clinEventCategoryConfirmed[!is.na(rhiccaDatWide$newOnsetMyocardialInfarction) & rhiccaDatWide$newOnsetMyocardialInfarction=="new MI"]<-paste(sep=";",rhiccaDatWide$clinEventCategoryConfirmed[!is.na(rhiccaDatWide$newOnsetMyocardialInfarction) & rhiccaDatWide$newOnsetMyocardialInfarction=="new MI"],"CVD")

for(j in 1:nrow(clinEventsDat)){
  pid<-clinEventsDat$PID[j]
  rhiccaDatWide$clinEventCategory[rhiccaDatWide$pid==pid]<-paste(sep=";",rhiccaDatWide$clinEventCategory[rhiccaDatWide$pid==pid],clinEventsDat$category[j])
  if(clinEventsDat$degreeOfCertainty[j]=="confirmed"){
    rhiccaDatWide$clinEventCategoryConfirmed[rhiccaDatWide$pid==pid]<-paste(sep=";",rhiccaDatWide$clinEventCategoryConfirmed[rhiccaDatWide$pid==pid],clinEventsDat$category[j])
    if(clinEventsDat$category[j]=="Systemic disease"){
      rhiccaDatWide$clinEventCategoryConfirmedSystemicType[rhiccaDatWide$pid==pid]<-paste(sep=";",rhiccaDatWide$clinEventCategoryConfirmedSystemicType[rhiccaDatWide$pid==pid],clinEventsDat$DIAGNOSIS[j])
    }
  }
}

rhiccaDatWide$clinEventCategory<-gsub(rhiccaDatWide$clinEventCategory,pattern="^;",replacement="")
rhiccaDatWide$clinEventCategoryConfirmed<-gsub(rhiccaDatWide$clinEventCategoryConfirmed,pattern="^;",replacement="")
rhiccaDatWide$clinEventCategoryConfirmedSystemicType<-gsub(rhiccaDatWide$clinEventCategoryConfirmedSystemicType,pattern="^;",replacement="")

rhiccaDatWide<-rhiccaDatWide %>%
  mutate(
    numClinEventsPerParticipant=0,
    numClinEventsPerParticipantNoOnset=0,
    numClinEventsPerParticipantNoPwvCimtAbpi=0,
    numClinEventsPerParticipantCVD=0,
    numClinEventsPerParticipantCVDNoOnset=0,
    numClinEventsPerParticipantInfection=0,
    numClinEventsPerParticipantSystemic=0,
    numClinEventsConfirmedPerParticipant=0,
    numClinEventsConfirmedPerParticipantWithDerivedWithoutPwvCimtAbpi=0,
    numClinEventsConfirmedPerParticipantCVD=0,
    numClinEventsConfirmedPerParticipantInfection=0,
    numClinEventsConfirmedPerParticipantSystemic=0,
    numClinEventsPwvCimtAbpi=0,
    numClinEventsPwvCimt=0,
    numClinEventsCimt=0
  )

idx<-which(rhiccaDatWide$clinEvent==1)

for(i in idx){
  rhiccaDatWide$numClinEventsPerParticipant[i]<-rhiccaDatWide$numClinEventsPerParticipant[i]+length(unlist(strsplit(split=";",rhiccaDatWide$clinEventCategory[i])))
  rhiccaDatWide$numClinEventsPerParticipantCVD[i]<-rhiccaDatWide$numClinEventsPerParticipantCVD[i]+sum(grepl(pattern="CVD",unlist(strsplit(split=";",rhiccaDatWide$clinEventCategory[i]))))
  rhiccaDatWide$numClinEventsPerParticipantInfection[i]<-rhiccaDatWide$numClinEventsPerParticipantInfection[i]+sum(grepl(pattern="Infection",unlist(strsplit(split=";",rhiccaDatWide$clinEventCategory[i]))))
  rhiccaDatWide$numClinEventsPerParticipantSystemic[i]<-rhiccaDatWide$numClinEventsPerParticipantSystemic[i]+sum(grepl(pattern="Systemic",unlist(strsplit(split=";",rhiccaDatWide$clinEventCategory[i]))))
  
  rhiccaDatWide$numClinEventsPerParticipantNoOnset[i]<-rhiccaDatWide$numClinEventsPerParticipant[i]-ifelse(!is.na(rhiccaDatWide$newOnsetCIMT[i]) & rhiccaDatWide$newOnsetCIMT[i]=="new CIMT vasculopathy",1,0)-ifelse(!is.na(rhiccaDatWide$newOnsetPWV[i]) & rhiccaDatWide$newOnsetPWV[i]=="new PWV vasculopathy",1,0)-ifelse(!is.na(rhiccaDatWide$newOnsetABPI[i]) & rhiccaDatWide$newOnsetABPI[i]=="new ABPI vasculopathy",1,0)-ifelse(!is.na(rhiccaDatWide$newOnsetDiabetes[i]) & rhiccaDatWide$newOnsetDiabetes[i]=="new diabetes",1,0)-ifelse(!is.na(rhiccaDatWide$newOnsetHypercholesterolemia[i]) & rhiccaDatWide$newOnsetHypercholesterolemia[i]=="new hypercholesterolemia",1,0)-ifelse(!is.na(rhiccaDatWide$newOnsetKidneyCondition[i]) & rhiccaDatWide$newOnsetKidneyCondition[i]=="new kidney condition",1,0)
  
  rhiccaDatWide$numClinEventsPerParticipantNoPwvCimtAbpi[i]<-rhiccaDatWide$numClinEventsPerParticipant[i]-ifelse(!is.na(rhiccaDatWide$newOnsetCIMT[i]) & rhiccaDatWide$newOnsetCIMT[i]=="new CIMT vasculopathy",1,0)-ifelse(!is.na(rhiccaDatWide$newOnsetPWV[i]) & rhiccaDatWide$newOnsetPWV[i]=="new PWV vasculopathy",1,0)-ifelse(!is.na(rhiccaDatWide$newOnsetABPI[i]) & rhiccaDatWide$newOnsetABPI[i]=="new ABPI vasculopathy",1,0)
  
  rhiccaDatWide$numClinEventsPerParticipantCVDNoOnset[i]<-rhiccaDatWide$numClinEventsPerParticipantCVD[i]-ifelse(!is.na(rhiccaDatWide$newOnsetCIMT[i]) & rhiccaDatWide$newOnsetCIMT[i]=="new CIMT vasculopathy",1,0)-ifelse(!is.na(rhiccaDatWide$newOnsetPWV[i]) & rhiccaDatWide$newOnsetPWV[i]=="new PWV vasculopathy",1,0)-ifelse(!is.na(rhiccaDatWide$newOnsetABPI[i]) & rhiccaDatWide$newOnsetABPI[i]=="new ABPI vasculopathy",1,0)
  
  rhiccaDatWide$numClinEventsPwvCimtAbpi[i]<-ifelse(!is.na(rhiccaDatWide$newOnsetPWV[i]) & rhiccaDatWide$newOnsetPWV[i]=="new PWV vasculopathy",1,0)+ifelse(!is.na(rhiccaDatWide$newOnsetCIMT[i]) & rhiccaDatWide$newOnsetCIMT[i]=="new CIMT vasculopathy",1,0)+ifelse(!is.na(rhiccaDatWide$newOnsetABPI[i]) & rhiccaDatWide$newOnsetABPI[i]=="new ABPI vasculopathy",1,0)
  
  rhiccaDatWide$numClinEventsPwvCimt[i]<-ifelse(!is.na(rhiccaDatWide$newOnsetPWV[i]) & rhiccaDatWide$newOnsetPWV[i]=="new PWV vasculopathy",1,0)+ifelse(!is.na(rhiccaDatWide$newOnsetCIMT[i]) & rhiccaDatWide$newOnsetCIMT[i]=="new CIMT vasculopathy",1,0)
  
  rhiccaDatWide$numClinEventsCimt[i]<-ifelse(!is.na(rhiccaDatWide$newOnsetCIMT[i]) & rhiccaDatWide$newOnsetCIMT[i]=="new CIMT vasculopathy",1,0)
  
  if(i %in% which(rhiccaDatWide$clinEventConfirmed==1)){
    rhiccaDatWide$numClinEventsConfirmedPerParticipant[i]<-rhiccaDatWide$numClinEventsConfirmedPerParticipant[i]+length(unlist(strsplit(split=";",rhiccaDatWide$clinEventCategoryConfirmed[i])))
    rhiccaDatWide$numClinEventsConfirmedPerParticipantWithDerivedWithoutPwvCimtAbpi[i]<-rhiccaDatWide$numClinEventsConfirmedPerParticipant[i]+length(unlist(strsplit(split=";",rhiccaDatWide$clinEventCategoryConfirmed[i])))+ifelse(!is.na(rhiccaDatWide$newOnsetDiabetes[i]) & rhiccaDatWide$newOnsetDiabetes[i]=="new diabetes",1,0)+ifelse(!is.na(rhiccaDatWide$newOnsetHypercholesterolemia[i]) & rhiccaDatWide$newOnsetHypercholesterolemia[i]=="new hypercholesterolemia",1,0)+ifelse(!is.na(rhiccaDatWide$newOnsetKidneyCondition[i]) & rhiccaDatWide$newOnsetKidneyCondition[i]=="new kidney condition",1,0)
    rhiccaDatWide$numClinEventsConfirmedPerParticipantCVD[i]<-rhiccaDatWide$numClinEventsConfirmedPerParticipantCVD[i]+sum(grepl(pattern="CVD",unlist(strsplit(split=";",rhiccaDatWide$clinEventCategoryConfirmed[i]))))
    rhiccaDatWide$numClinEventsConfirmedPerParticipantInfection[i]<-rhiccaDatWide$numClinEventsConfirmedPerParticipantInfection[i]+sum(grepl(pattern="Infection",unlist(strsplit(split=";",rhiccaDatWide$clinEventCategoryConfirmed[i]))))
    rhiccaDatWide$numClinEventsConfirmedPerParticipantSystemic[i]<-rhiccaDatWide$numClinEventsConfirmedPerParticipantSystemic[i]+sum(grepl(pattern="Systemic",unlist(strsplit(split=";",rhiccaDatWide$clinEventCategoryConfirmed[i]))))
  }
}

rhiccaDatWide$numClinEventsConfirmedPerParticipantCVDwithCIMT<-rhiccaDatWide$numClinEventsConfirmedPerParticipantCVD+ifelse(grepl(pattern="new",rhiccaDatWide$newOnsetCIMT),1,0)
rhiccaDatWide$numClinEventsConfirmedPerParticipantCVDwithCIMTPWV<-rhiccaDatWide$numClinEventsConfirmedPerParticipantCVDwithCIMT+ifelse(grepl(pattern="new",rhiccaDatWide$newOnsetPWV),1,0)

tmpDatAll <- rhiccaDatLong %>%
  dplyr::select(pid,crf.ptype,visit,pwv.pwv,cimt.cimt,abpi.abpi,eGFR,lims.CD4_Absolute) %>%
  filter(!is.na(crf.ptype) & is.element(visit,set=c("enrolment","6months","12months","18months","24months","30months","36months"))) %>%
  group_by(visit,crf.ptype) %>%
  summarise(pwvMean=mean(pwv.pwv,na.rm=T),pwvSE=sd(pwv.pwv,na.rm=T)/sqrt(sum(!is.na(pwv.pwv))),
            cimtMean=mean(cimt.cimt,na.rm=T),cimtSE=sd(cimt.cimt,na.rm=T)/sqrt(sum(!is.na(cimt.cimt))),
            abpiMean=mean(abpi.abpi,na.rm=T),abpiSE=sd(abpi.abpi,na.rm=T)/sqrt(sum(!is.na(abpi.abpi))),
            egfrMean=mean(eGFR,na.rm=T),egfrSE=sd(eGFR,na.rm=T)/sqrt(sum(!is.na(eGFR))),
            cd4Mean=mean(lims.CD4_Absolute,na.rm=T),cd4SE=sd(lims.CD4_Absolute,na.rm=T)/sqrt(sum(!is.na(lims.CD4_Absolute))),
            .groups="drop") %>%
  filter(visit!="unscheduled") %>%
  mutate(visitNum=ifelse(visit=="enrolment",0,as.integer(gsub(pattern="months",replacement="",visit))))

kableRn<-function(.data,row_names=NULL, ...){
  if(!is.null(row_names)){
    datTmp<-.data
    row.names(datTmp)<-row_names
    res<-kable(datTmp, row.names=T, ...)
  }else{
    res<-kable(.data, ...)
  }
  
  return(res)
}

idx<-which(!is.na(rhiccaDatWide$lims.Cholesterol_enrolment) & rhiccaDatWide$lims.Cholesterol_enrolment>=1000)
datTmp<-rhiccaDatWide[idx,]

rhiccaDatWide$lims.Cholesterol_enrolment[idx]<-NA

6.1.2 Loss to follow-up

The loss-to-follow-up rates over time are visualised on Figure 6.1.

binomCI<-function(x,n,type="low"){
  if(length(x)!=length(n)){stop("x and n need to be of same length!")}
  res<-rep(NA,length(x))
  
  for(i in 1:length(x)){
    if(type=="low"){
      res[i]<-binom.test(x=x[i],n=n[i])$conf.int[1]
    }else if(type=="high"){
      res[i]<-binom.test(x=x[i],n=n[i])$conf.int[2]
    }
  }
  
  return(res)
}

datTmp<-rhiccaDatWide %>%
  filter(ltfu_enrolment==0) %>%
  group_by(crf.ptype) %>%
  summarise(
    n=n(),
    n00=sum(ltfu_enrolment==1),
    n06=sum(ltfu_6months==1),
    n12=sum(ltfu_12months==1),
    n18=sum(ltfu_18months==1),
    n24=sum(ltfu_24months==1),
    .groups="drop"
  ) %>%
  pivot_longer(cols=c(n00,n06,n12,n18,n24),names_to="visit",values_to="k") %>%
  mutate(
    month=case_when(
      visit=="n00"~0,
      visit=="n06"~6,
      visit=="n12"~12,
      visit=="n18"~18,
      visit=="n24"~24
    ),
    prop=k/n,
    propLow=binomCI(x=k,n=n,type="low"),
    propHigh=binomCI(x=k,n=n,type="high")
  )

g<-datTmp %>%
  ggplot(mapping=aes(x=month,y=prop,ymin=propLow,ymax=propHigh,col=crf.ptype,fill=crf.ptype)) +
  geom_point(size=2) +
  geom_line(lwd=1) +
  geom_ribbon(alpha=0.25,linetype=0) +
  scale_color_manual(values=c("steelblue","orange"),name="HIV status") +
  scale_fill_manual(values=c("steelblue","orange"),name="HIV status") +
  guides(fill="none",lty="none",pch="none") +
  ggtitle("Loss-to-follow-up (LTFU) rates over time per participant type.") + 
  theme(text=element_text(size=16)) +
  xlab("Study visit (months since enrolment)") +
  ylab("LTFU proportion") 

print(g)

Figure 6.1: LTFU rates over time.

We can also tabulate the reasons for LTFU in Table 6.2.

datTmp<-rhiccaDatWide %>%
  filter(ltfu_enrolment==0) %>%
  group_by(crf.ptype) %>%
  summarise(
    n=n(),
    n24=sum(ltfu_24months==1),
    n24_death=sum(ltfu_24months==1 & ltfu_24months_reason=="DEATH",na.rm=T),
    n24_move=sum(ltfu_24months==1 & ltfu_24months_reason=="MIGRATED",na.rm=T),
    n24_consent=sum(ltfu_24months==1 & ltfu_24months_reason=="CONSENT WITHTDRAW",na.rm=T),
    n24_ineligible=sum(ltfu_24months==1 & ltfu_24months_reason=="RETROSPECTIVE DISCOVERY OF STUDY INELIGIBILITY",na.rm=T),
    n24_otherAlive=sum(ltfu_24months==1 & ltfu_24months_reason=="ALIVE",na.rm=T),
    n24_unknown=sum(ltfu_24months==1 & is.na(ltfu_24months_reason)),
    .groups="drop"
  )

cnames<-datTmp$crf.ptype
datTmp<-as.data.frame(t(datTmp[,-1]))
colnames(datTmp)<-cnames
rownames(datTmp)<-c("Total number of participants","Lost to follow-up","Dead","Moved away","Withdrew consent","Ineligible","Other reason but still alive","Unknown")

datTmp$ControlProp<-round(digits=3,100*datTmp$Control/datTmp$Control[1])
datTmp$PatientProp<-round(digits=3,100*datTmp$Patient/datTmp$Patient[1])

datTmp$Control<-paste(sep="",datTmp$Control," (",format(nsmall=2,datTmp$ControlProp),"%)")
datTmp$Patient<-paste(sep="",datTmp$Patient," (",format(nsmall=2,datTmp$PatientProp),"%)")

datTmp %>%
  dplyr::select(Control,Patient) %>%
  kable(caption="Reasons for loss-to-follow-up at 24 months by participant type.",align = "l") %>%
  kable_styling(full_width = F)

Table 6.2: Reasons for loss-to-follow-up at 24 months by participant type.
	Control	Patient
Total number of participants	190 (100.000%)	713 (100.000%)
Lost to follow-up	39 ( 20.526%)	193 ( 27.069%)
Dead	0 ( 0.000%)	37 ( 5.189%)
Moved away	7 ( 3.684%)	44 ( 6.171%)
Withdrew consent	4 ( 2.105%)	17 ( 2.384%)
Ineligible	4 ( 2.105%)	3 ( 0.421%)
Other reason but still alive	24 ( 12.632%)	86 ( 12.062%)
Unknown	0 ( 0.000%)	6 ( 0.842%)

datTmp<-rhiccaDatWide %>%
  filter(ltfu_enrolment==0 & ltfu_24months==1 & ltfu_24months_reason=="RETROSPECTIVE DISCOVERY OF STUDY INELIGIBILITY")

table(datTmp$crf.ptype,datTmp$ltfu_24months_ineligDetails) %>%
  t() %>%
  kable(caption="Reasons for retrospective ineligibility (participants with baseline CRF data only).") %>%
  kable_styling(full_width = F)

Table 6.2: Reasons for retrospective ineligibility (participants with baseline CRF data only).
	Control	Patient
NOT ART NAIIVE	0	3
PRE-EXISTING HEART FAILURE	1	0
SERO-CONVERSION	3	0

# removing participants without baseline CRF data
pidsNoCrf<-rhiccaDatWide$pid[rhiccaDatWide$ltfu_enrolment==1]
rhiccaDatWide<-rhiccaDatWide %>% filter(! pid %in% pidsNoCrf)
rhiccaDatLong<-rhiccaDatLong %>% filter(! pid %in% pidsNoCrf)

# Removing the ineligibles
pidsIneligible<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$ltfu_24months_reason) & rhiccaDatWide$ltfu_24months_reason=="RETROSPECTIVE DISCOVERY OF STUDY INELIGIBILITY"]
rhiccaDatWide<-rhiccaDatWide %>% filter(! pid %in% pidsIneligible)
rhiccaDatLong<-rhiccaDatLong %>% filter(! pid %in% pidsIneligible)

# cleaning redundant columns
sumNA<-colSums(is.na(rhiccaDatWide))
rhiccaDatWide<-rhiccaDatWide[,sumNA<nrow(rhiccaDatWide)]

obsTime<-rhiccaDatLong %>%
  dplyr::select(c(pid,crf.gender,crf.ptype,crf.age_group,contains("_date"))) %>%
  mutate(
    crf.data_date=dmy(crf.data_date),
    physio.data_date=dmy(physio.data_date),
    pwv.data_date=dmy(pwv.data_date),
    cimt.data_date=dmy(cimt.data_date),
    abpi.data_date=dmy(abpi.data_date),
    lims.data_date=as_date(mdy_hms(gsub(lims.sample_date,pattern="\\:000[AP]M",replacement="")))
  ) %>%
  mutate(
    lims.data_date=as_date(case_when(
      grepl(as.character(lims.data_date),pattern="2014")~NA_character_,
      TRUE~as.character(lims.data_date)
    ))
  ) %>%
  dplyr::select(!contains("sample")) %>%
  pivot_longer(cols=contains("_date"),names_to="type",values_to="date") %>%
  group_by(pid) %>%
  summarise(
    crf.gender=paste(collapse="_",unique(crf.gender)),
    crf.ptype=paste(collapse="_",unique(crf.ptype)),
    crf.age_group=paste(collapse="_",unique(crf.age_group)),
    recruitDate=min(date,na.rm=TRUE),
    lastVisitDate=max(date,na.rm=TRUE)
  ) %>%
  mutate(
    obsTime=ymd(lastVisitDate)-ymd(recruitDate)
  )

rhiccaDatWide<-rhiccaDatWide %>%
  mutate(
    recruitDate=as.character(obsTime$recruitDate[match(pid,obsTime$pid)]),
    lastVisitDate=as.character(obsTime$lastVisitDate[match(pid,obsTime$pid)]),
    obsTime=obsTime$obsTime[match(pid,obsTime$pid)]
    )

cmvvzvDat<-read.csv(cmvvzvFile)

# chat with Laura on 20/07/2022
# 1. For VZV ignore the confirmation results
# 2. For CMV, use the "CMV IgG Result" measurements unless these are ">250.0" in which case use the "CMV IgG Dilution" measurements
# Try if the missing Subject IDs can be inferred from the lab IDs ("Subject.ID.1")

cmvvzvDat$Subject.ID[cmvvzvDat$Subject.ID=="RE16903"]<-"RE1690J" # Laura indicated to correct this, but there aren't any instances where this applies
cmvvzvDat$Subject.ID[cmvvzvDat$Subject.ID=="RE1112B"]<-"RE1121B" # Laura indicated to correct this, but there aren't any instances where this applies
cmvvzvDat$Subject.ID[cmvvzvDat$Subject.ID=="RE1632Z"]<-"RE16322"
cmvvzvDat$Subject.ID[cmvvzvDat$Subject.ID=="RE1028S"]<-"RE10285"

cmvDat<-cmvvzvDat %>%
  dplyr::filter(Test=="CMV IgG Result")

cmvDilDat<-cmvvzvDat %>%
  dplyr::filter(Test=="CMV IgG Dilution")

vzvDat<-cmvvzvDat %>%
  dplyr::filter(Test=="Varizell Zoster IgG") 

for(j in 1:nrow(cmvDat)){
  if(cmvDat$Result[j]==">250.0"){
    #print(c(j,cmvDat$Subject.ID[j],cmvDat$Result[j],cmvDilDat$Result[cmvDilDat$labno==cmvDat$labno[j]]))
    if(length(cmvDilDat$Result[cmvDilDat$labno==cmvDat$labno[j]])!=0){
      cmvDat$Result[j]<-cmvDilDat$Result[cmvDilDat$labno==cmvDat$labno[j]]
    }
  }
}

idx2remove<-integer(0)
for(j in which(cmvDat$Subject.ID %in% c(".","RE"))){
  pid<-limsDatMerged$pid.x[limsDatMerged$lab_id==cmvDat$Subject.ID.1[j]]
  if(length(pid)==1){cmvDat$Subject.ID[j]<-pid}else{idx2remove<-c(idx2remove,j)}
}
if(length(idx2remove)>0){cmvDat<-cmvDat[-idx2remove,]}

idx2remove<-integer(0)
for(j in which(vzvDat$Subject.ID %in% c(".","RE"))){
  pid<-limsDatMerged$pid.x[limsDatMerged$lab_id==vzvDat$Subject.ID.1[j]]
  if(length(pid)==1){vzvDat$Subject.ID[j]<-pid}else{idx2remove<-c(idx2remove,j)}
}
if(length(idx2remove)>0){vzvDat<-vzvDat[-idx2remove,]}

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD10Q" & cmvDat$Subject.ID=="RE1869Y"]<-"CBD1Q0"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD10Q" & vzvDat$Subject.ID=="RE1869Y"]<-"CBD1Q0"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD114" & cmvDat$Subject.ID=="RE1539X"]<-"CBD1I4"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD114" & vzvDat$Subject.ID=="RE1539X"]<-"CBD1I4"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD115" & cmvDat$Subject.ID=="RE1540A"]<-"CBD1I5"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD115" & vzvDat$Subject.ID=="RE1540A"]<-"CBD1I5"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD11F" & cmvDat$Subject.ID=="RE15506"]<-"CBD1IF"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD11F" & vzvDat$Subject.ID=="RE15506"]<-"CBD1IF"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD11H" & cmvDat$Subject.ID=="RE1158M"]<-"CBD1IH"
cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBD1IH" & cmvDat$Subject.ID=="RE1158M"]<-"RE15522"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD11H" & vzvDat$Subject.ID=="RE1158M"]<-"CBD1IH"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBD1IH" & vzvDat$Subject.ID=="RE1158M"]<-"RE15522"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD11V" & cmvDat$Subject.ID=="RE15657"]<-"CBD1IV"
cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBD1IV" & cmvDat$Subject.ID=="RE15657"]<-"RE1565T"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD11V" & vzvDat$Subject.ID=="RE15657"]<-"CBD1IV"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBD1IV" & vzvDat$Subject.ID=="RE15657"]<-"RE1565T"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD11W" & cmvDat$Subject.ID=="RE1566R"]<-"CBD1IW"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD11W" & vzvDat$Subject.ID=="RE1566R"]<-"CBD1IW"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD12I" & cmvDat$Subject.ID=="RE1354E"]<-"CBE12I"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD12I" & vzvDat$Subject.ID=="RE1354E"]<-"CBE12I"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD12Y" & cmvDat$Subject.ID=="RE13754"]<-"CBE12Y"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD12Y" & vzvDat$Subject.ID=="RE13754"]<-"CBE12Y"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD14S" & cmvDat$Subject.ID=="RE15657"]<-"CBF14S"
cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBD14S" & cmvDat$Subject.ID=="RE1044Y"]<-"RE1144Y"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD14S" & vzvDat$Subject.ID=="RE15657"]<-"CBF14S"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBD14S" & vzvDat$Subject.ID=="RE1044Y"]<-"RE1144Y"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD17B" & cmvDat$Subject.ID=="RE1162N"]<-NA
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD17B" & vzvDat$Subject.ID=="RE1162N"]<-NA

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD188" & cmvDat$Subject.ID=="RE12610"]<-"CBF18B"
cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBD18B" & cmvDat$Subject.ID=="RE12610"]<-"RE1261O"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD188" & vzvDat$Subject.ID=="RE12610"]<-"CBF18B"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBD18B" & vzvDat$Subject.ID=="RE12610"]<-"RE1261O"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD18X" & cmvDat$Subject.ID %in% c("RE1249G","RE12634")]<-NA
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD18X" & vzvDat$Subject.ID %in% c("RE1249G","RE12634")]<-NA

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD1AQ" & cmvDat$Subject.ID=="RE1406M"]<-"CBD1A0"
cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBD1AQ" & cmvDat$Subject.ID=="RE1343M"]<-"RE1343K"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD1AQ" & vzvDat$Subject.ID=="RE1406M"]<-"CBD1A0"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBD1AQ" & vzvDat$Subject.ID=="RE1343M"]<-"RE1343K"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD1ED" & cmvDat$Subject.ID=="RE14582"]<-NA
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD1ED" & vzvDat$Subject.ID=="RE14582"]<-NA

cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBD1IK" & cmvDat$Subject.ID=="RE15542"]<-"RE1554Z"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBD1IK" & vzvDat$Subject.ID=="RE15542"]<-"RE1554Z"

cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBD1LT" & cmvDat$Subject.ID=="RE17142"]<-"RE1714Z"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBD1LT" & vzvDat$Subject.ID=="RE17142"]<-"RE1714Z"

cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBD1SD" & cmvDat$Subject.ID=="RE190D"]<-"RE1902R"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBD1SD" & vzvDat$Subject.ID=="RE190D"]<-"RE1902R"

cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBE114" & cmvDat$Subject.ID=="RE1828R"]<-"RE1828F"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBE114" & vzvDat$Subject.ID=="RE1828R"]<-"RE1828F"
cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBE114" & cmvDat$Subject.ID=="RE12424"]<-NA
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBE114" & vzvDat$Subject.ID=="RE12424"]<-NA

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBE115" & cmvDat$Subject.ID=="RE1250U"]<-"CBE11J"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBE115" & vzvDat$Subject.ID=="RE1250U"]<-"CBE11J"

cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBE11X" & cmvDat$Subject.ID=="RE12561"]<-"RE1256I"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBE11X" & vzvDat$Subject.ID=="RE12561"]<-"RE1256I"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF11J" & cmvDat$Subject.ID=="RE12117"]<-NA
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF11J" & vzvDat$Subject.ID=="RE12117"]<-NA

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF11S" & cmvDat$Subject.ID=="RE1218U"]<-NA
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF11S" & vzvDat$Subject.ID=="RE1218U"]<-NA

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF141" & cmvDat$Subject.ID=="RE1187C"]<-"CBF14I"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF141" & vzvDat$Subject.ID=="RE1187C"]<-"CBF14I"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF14U" & cmvDat$Subject.ID=="RE11173"]<-"CBF140"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF14U" & vzvDat$Subject.ID=="RE11173"]<-"CBF140"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF172" & cmvDat$Subject.ID=="RE1348A"]<-"CBF17Z"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF172" & vzvDat$Subject.ID=="RE1348A"]<-"CBF17Z"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF17F" & cmvDat$Subject.ID=="RE1414M"]<-"CBJ17F" # 24 months sample
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF17F" & vzvDat$Subject.ID=="RE1414M"]<-"CBJ17F"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF1ER" & cmvDat$Subject.ID=="RE11253"]<-"CBH1AR" # 12 months sample; wrong LIMS data year
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF1ER" & vzvDat$Subject.ID=="RE11253"]<-"CBH1AR"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF1EX" & cmvDat$Subject.ID=="RE1585C"]<-"CBF1FX"
cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBF1FX" & cmvDat$Subject.ID=="RE1585C"]<-"RE1585L"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF1EX" & vzvDat$Subject.ID=="RE1585C"]<-"CBF1FX"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBF1FX" & vzvDat$Subject.ID=="RE1585C"]<-"RE1585L"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF1EY" & cmvDat$Subject.ID=="RE1584N"]<-NA
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF1EY" & vzvDat$Subject.ID=="RE1584N"]<-NA

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF1GD" & cmvDat$Subject.ID=="RE1576N"]<-"CBF1GB"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF1GD" & vzvDat$Subject.ID=="RE1576N"]<-"CBF1GB"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF1H8" & cmvDat$Subject.ID=="RE1659H"]<-"CBF1HS"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF1H8" & vzvDat$Subject.ID=="RE1659H"]<-"CBF1HS"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF1HF" & cmvDat$Subject.ID=="RE1163S"]<-"CBH1HF" # 12 months sample
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF1HF" & vzvDat$Subject.ID=="RE1163S"]<-"CBH1HF"

cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBF1HF" & cmvDat$Subject.ID=="RE1693P"]<-"RE1693D"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBF1HF" & vzvDat$Subject.ID=="RE1693P"]<-"RE1693D"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF1I3" & cmvDat$Subject.ID=="RE15370"]<-NA
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF1I3" & vzvDat$Subject.ID=="RE15370"]<-NA

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD14S" & cmvDat$Subject.ID=="RE1185G"]<-"CBF14S"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD14S" & vzvDat$Subject.ID=="RE1185G"]<-"CBF14S"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD11B" & cmvDat$Subject.ID=="RE1041C"]<-NA
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD11B" & vzvDat$Subject.ID=="RE1041C"]<-NA

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF18B" & cmvDat$Subject.ID=="RE12610"]<-NA
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF18B" & vzvDat$Subject.ID=="RE12610"]<-NA

cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBD195" & cmvDat$Subject.ID=="RE12954"]<-"RE1292A"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBD195" & vzvDat$Subject.ID=="RE12954"]<-"RE1292A"

cmvDat$Subject.ID[cmvDat$Subject.ID.1=="CBF1KE" & cmvDat$Subject.ID=="RE1387X"]<-"RE1812V"
vzvDat$Subject.ID[vzvDat$Subject.ID.1=="CBF1KE" & vzvDat$Subject.ID=="RE1387X"]<-"RE1812V"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF16H" & cmvDat$Subject.ID=="RE1263K"]<-NA
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF16H" & vzvDat$Subject.ID=="RE1263K"]<-NA

# manual removal in line with Laura's email from 2022-08-03
labnos2Remove<-c("0022S082204","0022S081375","0022S079597","0022S081379")
cmvDat<-cmvDat %>% dplyr::filter(!(labno %in% labnos2Remove))
vzvDat<-vzvDat %>% dplyr::filter(!(labno %in% labnos2Remove))

problemlabIDscmv<-names(table(cmvDat$Subject.ID.1)[table(cmvDat$Subject.ID.1)>1]) # 12 left, down from 45 initially; many of these are proper duplicates but often with quite different measurements
problemlabIDsvzv<-names(table(vzvDat$Subject.ID.1)[table(vzvDat$Subject.ID.1)>1]) # 12 left, down from 45 initially; these many of these are proper duplicates but often with quite different measurements

for(ID in problemlabIDscmv){
  idx2remove<-character(0)
  if(sum(grepl(pattern="<|>",cmvDat$Result[cmvDat$Subject.ID.1==ID]))>0){
    idx2remove<-which(cmvDat$Subject.ID.1==ID & grepl(pattern="<|>",cmvDat$Result))
  }else if(sum(grepl(pattern="2022",cmvDat$Sample.Date[cmvDat$Subject.ID.1==ID]))>0){
    idx2remove<-which(cmvDat$Subject.ID.1==ID & !grepl(pattern="2022",cmvDat$Sample.Date))
  }
  if(length(idx2remove)>0){cmvDat<-cmvDat[-idx2remove,]}
}

for(ID in problemlabIDsvzv){
  idx2remove<-character(0)
  if(sum(grepl(pattern="<|>",vzvDat$Result[vzvDat$Subject.ID.1==ID]))>0){
    idx2remove<-which(vzvDat$Subject.ID.1==ID & grepl(pattern="<|>",vzvDat$Result))
  }else if(sum(grepl(pattern="2022",vzvDat$Sample.Date[vzvDat$Subject.ID.1==ID]))>0){
    idx2remove<-which(vzvDat$Subject.ID.1==ID & !grepl(pattern="2022",vzvDat$Sample.Date))
  }
  if(length(idx2remove)>0){vzvDat<-vzvDat[-idx2remove,]}
}

problemlabIDscmv<-names(table(cmvDat$Subject.ID.1)[table(cmvDat$Subject.ID.1)>1]) # 7 left, down from 45 initially; many of these are proper duplicates but often with quite different measurements
problemlabIDsvzv<-names(table(vzvDat$Subject.ID.1)[table(vzvDat$Subject.ID.1)>1]) # 7 left, down from 45 initially; these many of these are proper duplicates but often with quite different measurements

tmp.cmv<-cmvDat %>% dplyr::filter(Subject.ID.1 %in% problemlabIDscmv) 
tmp.vzv<-vzvDat %>% dplyr::filter(Subject.ID.1 %in% problemlabIDsvzv)
#tmp<-rbind(tmp.cmv,tmp.vzv)
#tmp<-tmp[order(tmp$Subject.ID.1),]
#write.csv(tmp,row.names=F,file=paste(sep="_",outPrefix,"CmvVzvProblemLabIDs.csv"))

# remove the duplicated results, before adding back the average values
cmvDat<-cmvDat %>% dplyr::filter(!(Subject.ID.1 %in% problemlabIDscmv))
vzvDat<-vzvDat %>% dplyr::filter(!(Subject.ID.1 %in% problemlabIDsvzv))

# average the duplicate results
tmp.cmv<-tmp.cmv %>%
  dplyr::group_by(Subject.ID) %>%
  dplyr::summarise(
    Source.Code=unique(Source.Code),
    labno=paste(collapse="_",labno),
    Subject.ID.1=unique(Subject.ID.1),
    Sample.Date=unique(Sample.Date),
    Test=unique(Test),
    Result=as.character(mean(as.numeric(Result))),
    Units=unique(Units),
    status=unique(status),
    .groups="drop"
  ) %>%
  dplyr::select(c(Source.Code,labno,Subject.ID,Subject.ID.1,Sample.Date,Test,Result,Units,status))

tmp.vzv<-tmp.vzv %>%
  dplyr::group_by(Subject.ID) %>%
  dplyr::summarise(
    Source.Code=unique(Source.Code),
    labno=paste(collapse="_",labno),
    Subject.ID.1=unique(Subject.ID.1),
    Sample.Date=unique(Sample.Date),
    Test=unique(Test),
    Result=as.character(mean(as.numeric(Result))),
    Units=unique(Units),
    status=unique(status),
    .groups="drop"
  ) %>%
  dplyr::select(c(Source.Code,labno,Subject.ID,Subject.ID.1,Sample.Date,Test,Result,Units,status))

cmvDat<-rbind(cmvDat,tmp.cmv)
vzvDat<-rbind(vzvDat,tmp.vzv)
rm(tmp.cmv,tmp.vzv)

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD15C" & cmvDat$Subject.ID=="RE1307S"]<-"CBD18X"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD15C" & vzvDat$Subject.ID=="RE1307S"]<-"CBD18X"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBE157" & cmvDat$Subject.ID=="RE1283C"]<-"CBD157"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBE157" & vzvDat$Subject.ID=="RE1283C"]<-"CBD157"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD1N2" & cmvDat$Subject.ID=="RE17935"]<-"CBD1N1"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD1N2" & vzvDat$Subject.ID=="RE17935"]<-"CBD1N1"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBD14P" & cmvDat$Subject.ID=="RE1156Q"]<-"CBF14P"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBD14P" & vzvDat$Subject.ID=="RE1156Q"]<-"CBF14P"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF1XO" & cmvDat$Subject.ID=="RE1968S"]<-"CBF1X0"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF1XO" & vzvDat$Subject.ID=="RE1968S"]<-"CBF1X0"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF1LO" & cmvDat$Subject.ID=="RE1850J"]<-"CBF1L0"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF1LO" & vzvDat$Subject.ID=="RE1850J"]<-"CBF1L0"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBG10W" & cmvDat$Subject.ID=="RE1305W"]<-"CBF17N"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBG10W" & vzvDat$Subject.ID=="RE1305W"]<-"CBF17N"

cmvDat$Subject.ID.1[cmvDat$Subject.ID.1=="CBF17N" & cmvDat$Subject.ID=="RE1307S"]<-"CBF16H"
vzvDat$Subject.ID.1[vzvDat$Subject.ID.1=="CBF17N" & vzvDat$Subject.ID=="RE1307S"]<-"CBF16H"

cmvDat<-cmvDat %>%
  dplyr::mutate(visit=rhiccaDatLong$visit[match(Subject.ID.1,rhiccaDatLong$lims.lab_id)]) %>%
  dplyr::mutate(visit=dplyr::case_when(
    is.na(Subject.ID.1)~NA_character_,
    TRUE~visit
    ))

vzvDat<-vzvDat %>%
  dplyr::mutate(visit=rhiccaDatLong$visit[match(Subject.ID.1,rhiccaDatLong$lims.lab_id)]) %>%
  dplyr::mutate(visit=dplyr::case_when(
    is.na(Subject.ID.1)~NA_character_,
    TRUE~visit
    ))

cmvDat$visit[cmvDat$Subject.ID.1=="CBG19Y" & cmvDat$Subject.ID=="RE1958W"]<-"6months"
vzvDat$visit[vzvDat$Subject.ID.1=="CBG19Y" & vzvDat$Subject.ID=="RE1958W"]<-"6months"

idxMissVisits<-which(is.na(cmvDat$visit) & cmvDat$Subject.ID %in% rhiccaDatLong$pid)
cmvDat$visit[idxMissVisits]<-rhiccaDatLong$visit[match(paste(sep="_",cmvDat$Subject.ID[idxMissVisits],lubridate::dmy(cmvDat$Sample.Date[idxMissVisits])),paste(sep="_",rhiccaDatLong$pid,lubridate::dmy(rhiccaDatLong$physio.data_date)))]

idxMissVisits<-which(is.na(vzvDat$visit) & vzvDat$Subject.ID %in% rhiccaDatLong$pid)
vzvDat$visit[idxMissVisits]<-rhiccaDatLong$visit[match(paste(sep="_",vzvDat$Subject.ID[idxMissVisits],lubridate::dmy(vzvDat$Sample.Date[idxMissVisits])),paste(sep="_",rhiccaDatLong$pid,lubridate::dmy(rhiccaDatLong$physio.data_date)))]

cmvDat<-cmvDat %>% dplyr::filter(Subject.ID %in% rhiccaDatWide$pid)
vzvDat<-vzvDat %>% dplyr::filter(Subject.ID %in% rhiccaDatWide$pid)

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE12858" & cmvDat$Subject.ID.1=="CBG10K"]<-"14/03/2018"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE12858" & vzvDat$Subject.ID.1=="CBG10K"]<-"14/03/2018"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE19711" & cmvDat$Subject.ID.1=="CBG19X"]<-"11/02/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE19711" & vzvDat$Subject.ID.1=="CBG19X"]<-"11/02/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE1930J" & cmvDat$Subject.ID.1=="CBG19W"]<-"08/02/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE1930J" & vzvDat$Subject.ID.1=="CBG19W"]<-"08/02/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE1922J" & cmvDat$Subject.ID.1=="CBG19S"]<-"06/02/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE1922J" & vzvDat$Subject.ID.1=="CBG19S"]<-"06/02/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE19287" & cmvDat$Subject.ID.1=="CBG19R"]<-"06/02/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE19287" & vzvDat$Subject.ID.1=="CBG19R"]<-"06/02/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE19279" & cmvDat$Subject.ID.1=="CBG19P"]<-"04/02/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE19279" & vzvDat$Subject.ID.1=="CBG19P"]<-"04/02/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE1926B" & cmvDat$Subject.ID.1=="CBG19N"]<-"01/02/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE1926B" & vzvDat$Subject.ID.1=="CBG19N"]<-"01/02/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE1870B" & cmvDat$Subject.ID.1=="CBG174"]<-"11/01/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE1870B" & vzvDat$Subject.ID.1=="CBG174"]<-"11/01/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE18735" & cmvDat$Subject.ID.1=="CBG11Y"]<-"11/01/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE18735" & vzvDat$Subject.ID.1=="CBG11Y"]<-"11/01/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE18663" & cmvDat$Subject.ID.1=="CBG171"]<-"10/01/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE18663" & vzvDat$Subject.ID.1=="CBG171"]<-"10/01/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE1854B" & cmvDat$Subject.ID.1=="CBG170"]<-"09/01/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE1854B" & vzvDat$Subject.ID.1=="CBG170"]<-"09/01/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE1868X" & cmvDat$Subject.ID.1=="CBG172"]<-"10/01/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE1868X" & vzvDat$Subject.ID.1=="CBG172"]<-"10/01/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE18671" & cmvDat$Subject.ID.1=="CBG173"]<-"10/01/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE18671" & vzvDat$Subject.ID.1=="CBG173"]<-"10/01/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE1853D" & cmvDat$Subject.ID.1=="CBG17N"]<-"07/01/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE1853D" & vzvDat$Subject.ID.1=="CBG17N"]<-"07/01/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE18567" & cmvDat$Subject.ID.1=="CBG17P"]<-"08/01/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE18567" & vzvDat$Subject.ID.1=="CBG17P"]<-"08/01/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE1830R" & cmvDat$Subject.ID.1=="CBG17Q"]<-"08/01/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE1830R" & vzvDat$Subject.ID.1=="CBG17Q"]<-"08/01/2019"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE1762J" & cmvDat$Subject.ID.1=="CBG17E"]<-"10/12/2018"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE1762J" & vzvDat$Subject.ID.1=="CBG17E"]<-"10/12/2018"

cmvDat$Sample.Date[cmvDat$Subject.ID=="RE18655" & cmvDat$Subject.ID.1=="CBG16Z"]<-"09/01/2019"
vzvDat$Sample.Date[vzvDat$Subject.ID=="RE18655" & vzvDat$Subject.ID.1=="CBG16Z"]<-"09/01/2019"

for(j in which(is.na(cmvDat$visit))){
  pid<-cmvDat$Subject.ID[j]
  sdate<-cmvDat$Sample.Date[j]
  j2<-which(rhiccaDatLong$pid==pid & abs(dmy(sdate)-dmy(rhiccaDatLong$physio.data_date))<5)
  j3<-which(vzvDat$Subject.ID==pid & vzvDat$Subject.ID.1==cmvDat$Subject.ID.1[j])
  if(length(j2)==1){
    cmvDat$visit[j]<-rhiccaDatLong$visit[j2]
    if(length(j3)==1){
      vzvDat$visit[j3]<-rhiccaDatLong$visit[j2]
    }
  }
}

colnames(cmvDat)[colnames(cmvDat)=="Result"]<-"cmv"
colnames(cmvDat)[colnames(cmvDat)=="Sample.Date"]<-"data_date"
cmvDat$visit<-factor(cmvDat$visit,levels=c("enrolment","6months","12months","18months","24months","30months","36months"))

colnames(vzvDat)[colnames(vzvDat)=="Result"]<-"vzv"
colnames(vzvDat)[colnames(vzvDat)=="Sample.Date"]<-"data_date"
vzvDat$visit<-factor(vzvDat$visit,levels=c("enrolment","6months","12months","18months","24months","30months","36months"))

cmvDat$cmv[cmvDat$cmv==">2500.0"]<-2500
cmvDat$cmv[cmvDat$cmv=="<0.1"]<-0.1
cmvDat$cmv<-as.numeric(cmvDat$cmv) # will lose one data point ">250.0"

vzvDat$vzv[vzvDat$vzv==">4000.0"]<-4000
vzvDat$vzv[vzvDat$vzv=="<10.0"]<-10.0
vzvDat$vzv<-as.numeric(vzvDat$vzv)

cmvDatWide<-cmvDat %>%
  dplyr::select(!c(Test,Units,status,labno)) %>%
  tidyr::pivot_wider(names_from=visit,values_from=c(cmv,data_date,Subject.ID.1))

vzvDatWide<-vzvDat %>%
  dplyr::select(!c(Test,Units,status,labno)) %>%
  tidyr::pivot_wider(names_from=visit,values_from=c(vzv,data_date,Subject.ID.1))

colnames(cmvDatWide)<-paste(sep=".","cmv",colnames(cmvDatWide))
colnames(vzvDatWide)<-paste(sep=".","vzv",colnames(vzvDatWide))

cmvQuartiles<-quantile(probs=c(0,0.25,0.5,0.75,1),c(cmvDatWide$cmv.cmv_enrolment,cmvDatWide$cmv.cmv_6months),na.rm=TRUE)
vzvQuartiles<-quantile(probs=c(0,0.25,0.5,0.75,1),c(vzvDatWide$vzv.vzv_enrolment,vzvDatWide$vzv.vzv_6months),na.rm=TRUE)

cmvDatWide<-cmvDatWide %>%
  dplyr::mutate(
    cmvStatus=factor(case_when(
      !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment<cmvQuartiles["75%"] & cmv.cmv_6months<cmvQuartiles["75%"] ~ "persistently low titres",
      !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment>=cmvQuartiles["75%"] & cmv.cmv_6months<cmvQuartiles["75%"] ~ "progression to low titres",
      !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment<cmvQuartiles["75%"] & cmv.cmv_6months>=cmvQuartiles["75%"] ~ "progression to high titres",
      !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment>=cmvQuartiles["75%"] & cmv.cmv_6months>=cmvQuartiles["75%"] ~ "persistently high titres",
      TRUE~NA_character_
    ),levels=c("persistently low titres","progression to low titres","progression to high titres","persistently high titres")),
    cmvEnrolmentStatus=factor(case_when(
      !is.na(cmv.cmv_enrolment) & cmv.cmv_enrolment<cmvQuartiles["75%"] ~ "low / medium enrolment titres",
      !is.na(cmv.cmv_enrolment) & cmv.cmv_enrolment>=cmvQuartiles["75%"] ~ "high enrolment titres",
      TRUE~NA_character_
    ),levels=c("low / medium enrolment titres","high enrolment titres"))
  ) %>%
  mutate(
        cmvStatus_2levels=factor(case_when(
          !is.na(cmvStatus) & cmvStatus=="persistently high titres"~"persistently high titres",
          !is.na(cmvStatus) & cmvStatus!="persistently high titres"~"not persistently high titres"
        ),levels=c("not persistently high titres","persistently high titres"))
  )

vzvDatWide<-vzvDatWide %>%
  dplyr::mutate(
    vzvStatus=factor(case_when(
      !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment<vzvQuartiles["75%"] & vzv.vzv_6months<vzvQuartiles["75%"] ~ "persistently low titres",
      !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment>=vzvQuartiles["75%"] & vzv.vzv_6months<vzvQuartiles["75%"] ~ "progression to low titres",
      !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment<vzvQuartiles["75%"] & vzv.vzv_6months>=vzvQuartiles["75%"] ~ "progression to high titres",
      !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment>=vzvQuartiles["75%"] & vzv.vzv_6months>=vzvQuartiles["75%"] ~ "persistently high titres",
      TRUE~NA_character_
    ),levels=c("persistently low titres","progression to low titres","progression to high titres","persistently high titres")),
    vzvEnrolmentStatus=factor(case_when(
      !is.na(vzv.vzv_enrolment) & vzv.vzv_enrolment<vzvQuartiles["75%"] ~ "low / medium enrolment titres",
      !is.na(vzv.vzv_enrolment) & vzv.vzv_enrolment>=vzvQuartiles["75%"] ~ "high enrolment titres",
      TRUE~NA_character_
    ),levels=c("low / medium enrolment titres","high enrolment titres"))
  ) %>%
  mutate(
        vzvStatus_2levels=factor(case_when(
          !is.na(vzvStatus) & vzvStatus=="persistently high titres"~"persistently high titres",
          !is.na(vzvStatus) & vzvStatus!="persistently high titres"~"not persistently high titres"
        ),levels=c("not persistently high titres","persistently high titres"))
  )

rhiccaDatWide<-left_join(rhiccaDatWide,cmvDatWide,by=c("pid"="cmv.Subject.ID"))
rhiccaDatWide<-left_join(rhiccaDatWide,vzvDatWide,by=c("pid"="vzv.Subject.ID"))

rhiccaDatLong<-rhiccaDatLong %>%
  dplyr::mutate(
    cmv.cmv=cmvDat$cmv[match(paste(sep="_",pid,visit),paste(sep="_",cmvDat$Subject.ID,cmvDat$visit))],
    vzv.vzv=vzvDat$vzv[match(paste(sep="_",pid,visit),paste(sep="_",vzvDat$Subject.ID,vzvDat$visit))]
  )

rhiccaDatWide<-rhiccaDatWide %>%
  dplyr::mutate(
    cmv.cmv_CFB=cmv.cmv_6months-cmv.cmv_enrolment,
    vzv.vzv_CFB=vzv.vzv_6months-vzv.vzv_enrolment,
    vlSuppression=case_when(
      lims.HIV_Abbott_Viral_Load_0pt2ml_P_12months<=1000~"Complete virological suppression",
      lims.HIV_Abbott_Viral_Load_0pt2ml_P_6months>1000 & lims.HIV_Abbott_Viral_Load_0pt2ml_P_12months>1000~"Incomplete virological suppression",
      lims.HIV_Abbott_Viral_Load_0pt2ml_P_6months<=1000 & lims.HIV_Abbott_Viral_Load_0pt2ml_P_12months>1000~"Virological resurgence"
    ),
    cd4Restoration=case_when(
      lims.CD4_Absolute_6months<200~"Minimal immune restoration",
      lims.CD4_Absolute_6months>=200 & lims.CD4_Absolute_6months<500~"Partial immune restoration",
      lims.CD4_Absolute_6months>=500~"Full immune restoration"
    )
  )

rhiccaDatWide<-rhiccaDatWide %>%
  dplyr::mutate(
    crf.PLWH=factor(case_when(crf.ptype=="Patient"~"PLWH",crf.ptype=="Control"~"Control")),
    framinghamScore=framinghamScore(rhiccaDatWide)
  )

rhiccaDatLong<-rhiccaDatLong %>%
  dplyr::mutate(
    crf.PLWH=rhiccaDatWide$crf.PLWH[match(pid,rhiccaDatWide$pid)],
    framinghamScore=rhiccaDatWide$framinghamScore[match(pid,rhiccaDatWide$pid)],
    physio.bmi_cat_enrolment=rhiccaDatWide$physio.bmi_cat_enrolment[match(pid,rhiccaDatWide$pid)],
    physio.hypertension_cat_enrolment=rhiccaDatWide$physio.bp_group_enrolment[match(pid,rhiccaDatWide$pid)],
    cmvStatus=rhiccaDatWide$cmvStatus[match(pid,rhiccaDatWide$pid)],
    vzvStatus=rhiccaDatWide$vzvStatus[match(pid,rhiccaDatWide$pid)],
    cmvStatus_2levels=rhiccaDatWide$cmvStatus_2levels[match(pid,rhiccaDatWide$pid)],
    vzvStatus_2levels=rhiccaDatWide$vzvStatus_2levels[match(pid,rhiccaDatWide$pid)],
    cmvEnrolmentStatus=rhiccaDatWide$cmvEnrolmentStatus[match(pid,rhiccaDatWide$pid)],
    vzvEnrolmentStatus=rhiccaDatWide$cmvEnrolmentStatus[match(pid,rhiccaDatWide$pid)],
    eGFR_existingCKD=rhiccaDatWide$eGFR_existingCKD[match(pid,rhiccaDatWide$pid)]
  )

# saving the cleaned, merged data
save(rhiccaDatLong,file=paste(sep="","../output/",gsub(pattern="-",replacement="",Sys.Date()),"/rhiccaDat_allMerged_longFormat",gsub(pattern="-",replacement="",Sys.Date()),".RData"))
save(rhiccaDatWide,file=paste(sep="","../output/",gsub(pattern="-",replacement="",Sys.Date()),"/rhiccaDat_allMerged_wideFormat",gsub(pattern="-",replacement="",Sys.Date()),".RData"))
write.csv(rhiccaDatWide,file=paste(sep="","../output/",gsub(pattern="-",replacement="",Sys.Date()),"/rhiccaDat_allMerged_wideFormat",gsub(pattern="-",replacement="",Sys.Date()),".csv"))

The final, cleaned data for analysis consists of 896 participants, of which 710 PLWH and 186 controls.

The total amount for follow-up time for all study participants was 1713.74 person-years, of which 1303.27 person-years were in PLWH and 410.48 in controls.

6.1.2.1 Study flow diagram

We can summarise the LTFU on a flow diagram, shown in Figure 6.2.

include_graphics("../SAP/ConsortDiagram_withTotals_20221109.png")

Figure 6.2: RHICCA study workflow.

6.1.3 Participant baseline data and demographics

We can summarise baseline data for most exposures, outcome data and demographics in Table 6.3.

sumFun<-function(.data){
  .data %>%
  dplyr::summarise(n=n(),
            # covariates confounders
            sexMale=sum(na.rm=T,crf.gender=="Male"),
            sexMaleProp=sum(na.rm=T,crf.gender=="Male")/sum(!is.na(crf.gender)),
            sexMaleMV=sum(is.na(crf.gender))/n(),
            ageMedian=median(na.rm=T,crf.age),
            age25th=quantile(na.rm=T,crf.age,probs=0.25),
            age75th=quantile(na.rm=T,crf.age,probs=0.75),
            ageMV=sum(is.na(crf.age))/n(),
            age3545=sum(na.rm=T,crf.age_group=="[35,45)"),
            age3545Prop=sum(na.rm=T,crf.age_group=="[35,45)")/sum(!is.na(crf.age_group)),
            age3545MV=NA,
            age4555=sum(na.rm=T,crf.age_group=="[45,55)"),
            age4555Prop=sum(na.rm=T,crf.age_group=="[45,55)")/sum(!is.na(crf.age_group)),
            age4555MV=NA,
            age5565=sum(na.rm=T,crf.age_group=="[55,65)"),
            age5565Prop=sum(na.rm=T,crf.age_group=="[55,65)")/sum(!is.na(crf.age_group)),
            age5565MV=NA,
            age65=sum(na.rm=T,crf.age_group=="[65,Inf)"),
            age65Prop=sum(na.rm=T,crf.age_group=="[65,Inf)")/sum(!is.na(crf.age_group)),
            age65MV=NA,
            BMIMedian=median(na.rm=T,physio.bmi_enrolment),
            BMI25th=quantile(na.rm=T,physio.bmi_enrolment,probs=0.25),
            BMI75th=quantile(na.rm=T,physio.bmi_enrolment,probs=0.75),
            BMIMV=sum(is.na(physio.bmi_enrolment))/n(),
            WHRMedian=median(na.rm=T,physio.waist_enrolment/physio.hip_enrolment),
            WHR25th=quantile(na.rm=T,physio.waist_enrolment/physio.hip_enrolment,probs=0.25),
            WHR75th=quantile(na.rm=T,physio.waist_enrolment/physio.hip_enrolment,probs=0.75),
            WHRMV=sum(is.na(physio.waist_enrolment/physio.hip_enrolment))/n(),
            bpDiaMean=mean(na.rm=T,physio.bp_dia_enrolment),
            bpDiaSd=sd(na.rm=T,physio.bp_dia_enrolment),
            bpDiaMV=sum(is.na(physio.bp_dia_enrolment))/n(),
            bpSysMean=mean(na.rm=T,physio.bp_sys_enrolment),
            bpSysSd=sd(na.rm=T,physio.bp_sys_enrolment),
            bpSysMV=sum(is.na(physio.bp_sys_enrolment))/n(),
            tensionNormal=sum(na.rm=T,physio.bp_group_enrolment=="Normotensive"),
            tensionNormalProp=sum(na.rm=T,physio.bp_group_enrolment=="Normotensive")/sum(!is.na(physio.bp_group_enrolment)),
            tensionNormalMV=NA,
            tensionPreHyp=sum(na.rm=T,physio.bp_group_enrolment=="Pre-hypertensive"),
            tensionPreHypProp=sum(na.rm=T,physio.bp_group_enrolment=="Pre-hypertensive")/sum(!is.na(physio.bp_group_enrolment)),
            tensionPreHypMV=NA,
            tensionHyp=sum(na.rm=T,physio.bp_group_enrolment=="Hypertensive"),
            tensionHypProp=sum(na.rm=T,physio.bp_group_enrolment=="Hypertensive")/sum(!is.na(physio.bp_group_enrolment)),
            tensionHypMV=NA,
            cholesterolMedian=median(na.rm=T,lims.Cholesterol_enrolment),
            cholesterol25th=quantile(na.rm=T,lims.Cholesterol_enrolment,probs=0.25),
            cholesterol75th=quantile(na.rm=T,lims.Cholesterol_enrolment,probs=0.75),
            cholesterolMV=sum(is.na(lims.Cholesterol_enrolment))/n(),
            hyperchol=sum(na.rm=T,lims.Cholesterol_enrolment>=5.2),
            hypercholProp=sum(na.rm=T,lims.Cholesterol_enrolment>=5.2)/sum(!is.na(lims.Cholesterol_enrolment)),
            hypercholMV=NA,
            glucoseMean=mean(na.rm=T,lims.Glucose_enrolment),
            glucoseSd=sd(na.rm=T,lims.Glucose_enrolment),
            glucoseMV=sum(is.na(lims.Glucose_enrolment))/n(),
            diabetes=sum(na.rm=T,lims.Glucose_enrolment>=11.1),
            diabetesProp=sum(na.rm=T,lims.Glucose_enrolment>=11.1)/sum(!is.na(lims.Glucose_enrolment)),
            diabetesMV=NA,
            eGFRMean=mean(na.rm=T,eGFR_enrolment),
            eGFRSd=sd(na.rm=T,eGFR_enrolment),
            eGFRMV=sum(is.na(eGFR_enrolment))/n(),
            chronKidney=sum(na.rm=T,eGFR_enrolment<60),
            chronKidneyProp=sum(na.rm=T,eGFR_enrolment<60)/sum(!is.na(eGFR_enrolment)),
            chronKidneyMV=NA,
            curSmoke=sum(na.rm=T,crf.smoke=="yes"),
            curSmokeProp=sum(na.rm=T,crf.smoke=="yes")/sum(!is.na(crf.smoke)),
            curSmokeMV=sum(is.na(crf.smoke))/n(),
            alcohol=sum(na.rm=T,crf.alcohol=="yes"),
            alcoholProp=sum(na.rm=T,crf.alcohol=="yes")/sum(!is.na(crf.alcohol)),
            alcoholMV=sum(is.na(crf.alcohol))/n(),
            framHamScoreMedian=median(na.rm=T,framinghamScore),
            framHamScore25th=quantile(na.rm=T,framinghamScore,probs=0.25),
            framHamScore75th=quantile(na.rm=T,framinghamScore,probs=0.75),
            framHamScoreMV=sum(is.na(framinghamScore))/n(),
            
            # covariates primary
            HIVpos=sum(crf.ptype=="Patient"),
            HIVposProp=sum(crf.ptype=="Patient")/sum(!is.na(crf.ptype)),
            HIVposMV=sum(is.na(crf.ptype))/n(),
            HIVvlMedian=median(na.rm=T,lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment),
            HIVvl25th=quantile(na.rm=T,lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment,probs=0.25),
            HIVvl75th=quantile(na.rm=T,lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment,probs=0.75),
            HIVvlMV=sum(is.na(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment))/n(),
            HIVvl_6mts_Median=median(na.rm=T,lims.HIV_Abbott_Viral_Load_0pt2ml_P_6months),
            HIVvl_6mts_25th=quantile(na.rm=T,lims.HIV_Abbott_Viral_Load_0pt2ml_P_6months,probs=0.25),
            HIVvl_6mts_75th=quantile(na.rm=T,lims.HIV_Abbott_Viral_Load_0pt2ml_P_6months,probs=0.75),
            HIVvl_6mts_MV=sum(is.na(lims.HIV_Abbott_Viral_Load_0pt2ml_P_6months))/n(),
            CD4Median=median(na.rm=T,lims.CD4_Absolute_enrolment),
            CD425th=quantile(na.rm=T,lims.CD4_Absolute_enrolment,probs=0.25),
            CD475th=quantile(na.rm=T,lims.CD4_Absolute_enrolment,probs=0.75),
            CD4MV=sum(is.na(lims.CD4_Absolute_enrolment))/n(),
            CD4_6mts_Median=median(na.rm=T,lims.CD4_Absolute_6months),
            CD4_6mts_25th=quantile(na.rm=T,lims.CD4_Absolute_6months,probs=0.25),
            CD4_6mts_75th=quantile(na.rm=T,lims.CD4_Absolute_6months,probs=0.75),
            CD4_6mts_MV=sum(is.na(lims.CD4_Absolute_6months))/n(),
            CD163Median=median(na.rm=T,lum.CD163_enrolment),
            CD16325th=quantile(na.rm=T,lum.CD163_enrolment,probs=0.25),
            CD16375th=quantile(na.rm=T,lum.CD163_enrolment,probs=0.75),
            CD163MV=sum(is.na(lum.CD163_enrolment))/n(),
            ICAM1Median=median(na.rm=T,lum.ICAM1_enrolment),
            ICAM125th=quantile(na.rm=T,lum.ICAM1_enrolment,probs=0.25),
            ICAM175th=quantile(na.rm=T,lum.ICAM1_enrolment,probs=0.75),
            ICAM1MV=sum(is.na(lum.ICAM1_enrolment))/n(),
            IL6Median=median(na.rm=T,lum.IL6_enrolment),
            IL625th=quantile(na.rm=T,lum.IL6_enrolment,probs=0.25),
            IL675th=quantile(na.rm=T,lum.IL6_enrolment,probs=0.75),
            IL6MV=sum(is.na(lum.IL6_enrolment))/n(),
            IL1bMedian=median(na.rm=T,lum.ILb_enrolment),
            IL1b25th=quantile(na.rm=T,lum.ILb_enrolment,probs=0.25),
            IL1b75th=quantile(na.rm=T,lum.ILb_enrolment,probs=0.75),
            IL1bMV=sum(is.na(lum.ILb_enrolment))/n(),
            cmvMedian=median(na.rm=T,cmv.cmv_enrolment),
            cmv25th=quantile(na.rm=T,cmv.cmv_enrolment,probs=0.25),
            cmv75th=quantile(na.rm=T,cmv.cmv_enrolment,probs=0.75),
            cmvMV=sum(is.na(cmv.cmv_enrolment))/n(),
            cmvStatusMV=sum(is.na(cmv.cmv_enrolment) | is.na(cmv.cmv_6months))/n(),
            cmvStatusPLow=sum(na.rm=TRUE,cmvStatus=="persistently low titres"),
            cmvStatusPLowProp=sum(na.rm=TRUE,cmvStatus=="persistently low titres")/sum(!is.na(cmvStatus)),
            cmvStatusPLowMV=NA,
            cmvStatusPLat=sum(na.rm=TRUE,cmvStatus=="progression to low titres"),
            cmvStatusPLatProp=sum(na.rm=TRUE,cmvStatus=="progression to low titres")/sum(!is.na(cmvStatus)),
            cmvStatusPLatMV=NA,
            cmvStatusRLat=sum(na.rm=TRUE,cmvStatus=="progression to high titres"),
            cmvStatusRLatProp=sum(na.rm=TRUE,cmvStatus=="progression to high titres")/sum(!is.na(cmvStatus)),
            cmvStatusRLatMV=NA,
            cmvStatusPHigh=sum(na.rm=TRUE,cmvStatus=="persistently high titres"),
            cmvStatusPHighProp=sum(na.rm=TRUE,cmvStatus=="persistently high titres")/sum(!is.na(cmvStatus)),
            cmvStatusPHighMV=NA,
            vzvMedian=median(na.rm=T,vzv.vzv_enrolment),
            vzv25th=quantile(na.rm=T,vzv.vzv_enrolment,probs=0.25),
            vzv75th=quantile(na.rm=T,vzv.vzv_enrolment,probs=0.75),
            vzvMV=sum(is.na(vzv.vzv_enrolment))/n(),
            vzvStatusMV=sum(is.na(vzv.vzv_enrolment) | is.na(vzv.vzv_6months))/n(),
            vzvStatusPLow=sum(na.rm=TRUE,vzvStatus=="persistently low titres"),
            vzvStatusPLowProp=sum(na.rm=TRUE,vzvStatus=="persistently low titres")/sum(!is.na(vzvStatus)),
            vzvStatusPLowMV=NA,
            vzvStatusPLat=sum(na.rm=TRUE,vzvStatus=="progression to low titres"),
            vzvStatusPLatProp=sum(na.rm=TRUE,vzvStatus=="progression to low titres")/sum(!is.na(vzvStatus)),
            vzvStatusPLatMV=NA,
            vzvStatusRLat=sum(na.rm=TRUE,vzvStatus=="progression to high titres"),
            vzvStatusRLatProp=sum(na.rm=TRUE,vzvStatus=="progression to high titres")/sum(!is.na(cmvStatus)),
            vzvStatusRLatMV=NA,
            vzvStatusPHigh=sum(na.rm=TRUE,vzvStatus=="persistently high titres"),
            vzvStatusPHighProp=sum(na.rm=TRUE,vzvStatus=="persistently high titres")/sum(!is.na(vzvStatus)),
            vzvStatusPHighMV=NA,
            
            # endpoints
            PWVMean=mean(na.rm=T,pwv.pwv_enrolment),
            PWVSd=sd(na.rm=T,pwv.pwv_enrolment),
            PWVMV=sum(is.na(pwv.pwv_enrolment))/n(),
            CIMTMean=mean(na.rm=T,cimt.cimt_enrolment),
            CIMTSd=sd(na.rm=T,cimt.cimt_enrolment),
            CIMTMV=sum(is.na(cimt.cimt_enrolment))/n(),
            ABPIMean=mean(na.rm=T,abpi.abpi_enrolment),
            ABPISd=sd(na.rm=T,abpi.abpi_enrolment),
            ABPIMV=sum(is.na(abpi.abpi_enrolment))/n(),
            clinEvents=sum(na.rm=T,numClinEventsPerParticipant),
            clinEventsPer100Participants=100*sum(na.rm=T,numClinEventsPerParticipant)/sum(!is.na(numClinEventsPerParticipant)),
            clinEventsMV=sum(is.na(numClinEventsPerParticipant))/n(),
            clinEventsConfirmedCVD=sum(na.rm=T,numClinEventsConfirmedPerParticipantCVD),
            clinEventsConfirmedCVDPer1000PYs=1000*365.25*sum(na.rm=T,numClinEventsConfirmedPerParticipantCVD)/sum(na.rm=T,as.numeric(obsTime)),
            clinEventsConfirmedCVDMV=sum(is.na(numClinEventsConfirmedPerParticipant))/n(),
            clinEventsConfirmedCVDWithCIMT=sum(na.rm=T,numClinEventsConfirmedPerParticipantCVDwithCIMT),
            clinEventsConfirmedCVDWithCIMTPer1000PYs=1000*365.25*sum(na.rm=T,numClinEventsConfirmedPerParticipantCVDwithCIMT)/sum(na.rm=T,as.numeric(obsTime)),
            clinEventsConfirmedCVDWithCIMTMV=sum(is.na(numClinEventsConfirmedPerParticipantCVDwithCIMT))/n(),
            clinEventsCimt=sum(na.rm=T,numClinEventsCimt),
            clinEventsCimtPer1000PYs=1000*365.25*sum(na.rm=T,numClinEventsCimt)/sum(na.rm=T,as.numeric(obsTime)),
            clinEventsCimtMV=sum(is.na(numClinEventsCimt))/n(),
            
            # follow-up
            followUp24mts=sum(ltfu_24months==0),
            followUp24mtsProp=sum(ltfu_24months==0)/n(),
            followUp24mtsMV=NA,
            .groups="drop")
}

demoDatAll<-rhiccaDatWide %>%
  filter(ltfu_enrolment==0) %>%
  sumFun()

demoDatStrat<-rhiccaDatWide %>%
  filter(ltfu_enrolment==0) %>%
  group_by(crf.ptype) %>%
  sumFun()

demoDatSummary<-rbind(data.frame(crf.ptype="All",demoDatAll),demoDatStrat)


reformatFun<-function(dat,var,type,rowIdx,roundDigits=NULL){
  # type needs to be one of "medianIQR", "meanSD", "nProp"
  if(type=="medianIQR"){
    if(is.null(roundDigits)){roundDigits=1}
    tmp<-as.numeric(dat[rowIdx,paste(sep="",var,c("MV","Median","25th","75th"))]) 
    tmp<-c(paste(sep="",format(nsmall=1,round(100*tmp[1],digits=1)),"%"),format(nsmall=roundDigits,round(tmp[2],digits=roundDigits)),paste(sep="","(",format(nsmall=roundDigits,round(tmp[3],digits=roundDigits)),"-",format(nsmall=roundDigits,round(tmp[4],digits=roundDigits)),")"))
  }else if(type=="medianIQRbignum"){
    if(is.null(roundDigits)){roundDigits=1}
    tmp<-as.numeric(dat[rowIdx,paste(sep="",var,c("MV","Median","25th","75th"))]) 
    tmp<-c(paste(sep="",format(nsmall=1,round(100*tmp[1],digits=1)),"%"),format(big.mark=",",nsmall=roundDigits,round(tmp[2],digits=roundDigits)),paste(sep="","(",format(big.mark=",",nsmall=roundDigits,round(tmp[3],digits=roundDigits)),"-",format(big.mark=",",nsmall=roundDigits,round(tmp[4],digits=roundDigits)),")"))
  }else if(type=="medianIQRperc"){
    if(is.null(roundDigits)){roundDigits=1}
    tmp<-as.numeric(dat[rowIdx,paste(sep="",var,c("MV","Median","25th","75th"))]) 
    tmp<-c(paste(sep="",format(nsmall=1,round(100*tmp[1],digits=1)),"%"),paste(sep="",format(nsmall=roundDigits,round(100*tmp[2],digits=roundDigits)),"%"),paste(sep="","(",paste(sep="",format(nsmall=roundDigits,round(100*tmp[3],digits=roundDigits)),"%"),"-",paste(sep="",format(nsmall=roundDigits,round(100*tmp[4],digits=roundDigits)),"%"),")"))
  }else if(type=="meanSD"){
    if(is.null(roundDigits)){roundDigits=1}
    tmp<-as.numeric(dat[rowIdx,paste(sep="",var,c("MV","Mean","Sd"))])
    tmp<-c(paste(sep="",format(nsmall=1,round(100*tmp[1],digits=1)),"%"),format(nsmall=roundDigits,round(tmp[2],digits=roundDigits)),paste(sep="","(",format(nsmall=roundDigits,round(tmp[3],digits=roundDigits)),")"))
  }else if(type=="nProp"){
    if(is.null(roundDigits)){roundDigits=1}
    tmp<-as.numeric(dat[rowIdx,paste(sep="",var,c("MV","","Prop"))])
    tmp<-c(paste(sep="",format(nsmall=1,round(100*tmp[1],digits=1)),"%"),tmp[2],paste(sep="","(",format(nsmall=roundDigits,round(100*tmp[3],digits=roundDigits)),"%)"))
  }
  
  tmp[tmp=="NA%"]<-NA
  
  return(tmp)
}

demoDatKable<-data.frame(
  variable=c("Participants","Followed-up at 24 months","Sex: male","Age","Age group [35,45)","Age group [45,55)","Age group [55,65)","Age group 65+","BMI (kg/m^2^)","Waist-hip ratio","Diastolic blood pressure (mm Hg)","Systolic blood pressure (mm Hg)","Blood pressure: normotensive","Blood pressure: pre-hypertensive","Blood pressure: hypertensive","Cholesterol (mmol/l)","Hypercholesteremic","Glucose (mmol/l)","Diabetic","eGFR (ml/min/1.73m^2^)","Chronic kidney disease","Current smoker","Drink alcohol","Framingham 10-year CVD risk","HIV infected","HIV viral load at baseline (copies/ml)","HIV viral load at 6 months (copies/ml)","HIV CD4+ at baseline (cells/mm^3^)","HIV CD4+ at 6 months (cells/mm^3^)","CD163 (ng/ml)","ICAM-1 (pg/ml)","IL-6 (pg/ml)","IL-1b (pg/ml)","CMV (AU/ml)","CMV status","CMV status: persistently low titres","CMV status: progression to low titres","CMV status: progression to high titres","CMV status: persistently high titres","VZV (mIU/ml)","VZV status","VZV status: persistently low titres","VZV status: progression to low titres","VZV status: progression to high titres","VZV status: persistently high titres","PWV (m/s)","CIMT (mm)","ABPI","Confirmed clinical CVD/CBD events","Sub-clinical CIMT events","Sub-/clinical CVD/CBD events"),
  statistics=c("n (% out of total)","n (%)","n (%)","median (IQR)","n (%)","n (%)","n (%)","n (%)","median (IQR)","median (IQR)","mean (SD)","mean (SD)","n (%)","n (%)","n (%)","median (IQR)","n (%)","mean (SD)","n (%)","mean (SD)","n (%)","n (%)","n (%)","median (IQR)","n (%)","median (IQR)","median (IQR)","median (IQR)","median (IQR)","median (IQR)","median (IQR)","median (IQR)","median (IQR)","median (IQR)","","n (%)","n (%)","n (%)","n (%)","median (IQR)","","n (%)","n (%)","n (%)","n (%)","mean (SD)","mean (SD)","mean (SD)","n (events per 1,000 person-years)","n (events per 1,000 person-years)","n (events per 1,000 person-years)"),
  AMV=NA,
  Astat1=NA,
  Astat2=NA,
  PMV=NA,
  Pstat1=NA,
  Pstat2=NA,
  CNA=NA,
  Cstat1=NA,
  Cstat2=NA
)

demoDatKable[1,-c(1,2)]<-c(NA,sum(rhiccaDatWide$ltfu_enrolment==0 & !is.na(rhiccaDatWide$crf.ptype)),
                           paste(sep="","(",format(nsmall=1,round(100*sum(rhiccaDatWide$ltfu_enrolment==0 & !is.na(rhiccaDatWide$crf.ptype))/sum(rhiccaDatWide$ltfu_enrolment==0 & !is.na(rhiccaDatWide$crf.ptype)),digits=1)),"%)"),
                           NA,sum(rhiccaDatWide$ltfu_enrolment==0 & !is.na(rhiccaDatWide$crf.ptype) & rhiccaDatWide$crf.ptype=="Patient"),
                           paste(sep="","(",format(nsmall=1,round(100*sum(rhiccaDatWide$ltfu_enrolment==0 & !is.na(rhiccaDatWide$crf.ptype) & rhiccaDatWide$crf.ptype=="Patient")/sum(rhiccaDatWide$ltfu_enrolment==0 & !is.na(rhiccaDatWide$crf.ptype)),digits=1)),"%)"),
                           NA,sum(rhiccaDatWide$ltfu_enrolment==0 & !is.na(rhiccaDatWide$crf.ptype) & rhiccaDatWide$crf.ptype=="Control"),
                           paste(sep="","(",format(nsmall=1,round(100*sum(rhiccaDatWide$ltfu_enrolment==0 & !is.na(rhiccaDatWide$crf.ptype) & rhiccaDatWide$crf.ptype=="Control")/sum(rhiccaDatWide$ltfu_enrolment==0 & !is.na(rhiccaDatWide$crf.ptype)),digits=1)),"%)"))

demoDatKable[demoDatKable$variable=="Followed-up at 24 months",-c(1,2)]<-c(reformatFun(dat=demoDatSummary,var="followUp24mts",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="followUp24mts",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="followUp24mts",type="nProp",rowIdx=2))

demoDatKable[demoDatKable$variable=="Sex: male",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="sexMale",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="sexMale",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="sexMale",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="Age",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="age",type="medianIQR",rowIdx=1),reformatFun(dat=demoDatSummary,var="age",type="medianIQR",rowIdx=3),reformatFun(dat=demoDatSummary,var="age",type="medianIQR",rowIdx=2))
demoDatKable[demoDatKable$variable=="Age group [35,45)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="age3545",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="age3545",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="age3545",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="Age group [45,55)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="age4555",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="age4555",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="age4555",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="Age group [55,65)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="age5565",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="age5565",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="age5565",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="Age group 65+",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="age65",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="age65",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="age65",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="BMI (kg/m^2^)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="BMI",type="medianIQR",rowIdx=1),reformatFun(dat=demoDatSummary,var="BMI",type="medianIQR",rowIdx=3),reformatFun(dat=demoDatSummary,var="BMI",type="medianIQR",rowIdx=2))
demoDatKable[demoDatKable$variable=="Waist-hip ratio",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="WHR",type="medianIQR",rowIdx=1),reformatFun(dat=demoDatSummary,var="WHR",type="medianIQR",rowIdx=3),reformatFun(dat=demoDatSummary,var="WHR",type="medianIQR",rowIdx=2))
demoDatKable[demoDatKable$variable=="Diastolic blood pressure (mm Hg)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="bpDia",type="meanSD",rowIdx=1),reformatFun(dat=demoDatSummary,var="bpDia",type="meanSD",rowIdx=3),reformatFun(dat=demoDatSummary,var="bpDia",type="meanSD",rowIdx=2))
demoDatKable[demoDatKable$variable=="Systolic blood pressure (mm Hg)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="bpSys",type="meanSD",rowIdx=1),reformatFun(dat=demoDatSummary,var="bpSys",type="meanSD",rowIdx=3),reformatFun(dat=demoDatSummary,var="bpSys",type="meanSD",rowIdx=2))
demoDatKable[demoDatKable$variable=="Blood pressure: normotensive",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="tensionNormal",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="tensionNormal",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="tensionNormal",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="Blood pressure: pre-hypertensive",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="tensionPreHyp",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="tensionPreHyp",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="tensionPreHyp",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="Blood pressure: hypertensive",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="tensionHyp",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="tensionHyp",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="tensionHyp",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="Cholesterol (mmol/l)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="cholesterol",type="medianIQR",rowIdx=1),reformatFun(dat=demoDatSummary,var="cholesterol",type="medianIQR",rowIdx=3),reformatFun(dat=demoDatSummary,var="cholesterol",type="medianIQR",rowIdx=2))
demoDatKable[demoDatKable$variable=="Hypercholesteremic",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="hyperchol",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="hyperchol",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="hyperchol",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="Glucose (mmol/l)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="glucose",type="meanSD",rowIdx=1),reformatFun(dat=demoDatSummary,var="glucose",type="meanSD",rowIdx=3),reformatFun(dat=demoDatSummary,var="glucose",type="meanSD",rowIdx=2))
demoDatKable[demoDatKable$variable=="Diabetic",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="diabetes",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="diabetes",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="diabetes",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="eGFR (ml/min/1.73m^2^)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="eGFR",type="meanSD",rowIdx=1),reformatFun(dat=demoDatSummary,var="eGFR",type="meanSD",rowIdx=3),reformatFun(dat=demoDatSummary,var="eGFR",type="meanSD",rowIdx=2))
demoDatKable[demoDatKable$variable=="Chronic kidney disease",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="chronKidney",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="chronKidney",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="chronKidney",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="Current smoker",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="curSmoke",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="curSmoke",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="curSmoke",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="Drink alcohol",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="alcohol",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="alcohol",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="alcohol",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="Framingham 10-year CVD risk",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="framHamScore",type="medianIQRperc",rowIdx=1),reformatFun(dat=demoDatSummary,var="framHamScore",type="medianIQRperc",rowIdx=3),reformatFun(dat=demoDatSummary,var="framHamScore",type="medianIQRperc",rowIdx=2))
demoDatKable[demoDatKable$variable=="HIV infected",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="HIVpos",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="HIVpos",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="HIVpos",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="HIV viral load at baseline (copies/ml)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="HIVvl",type="medianIQRbignum",rowIdx=1,roundDigits=0),reformatFun(dat=demoDatSummary,var="HIVvl",type="medianIQRbignum",rowIdx=3,roundDigits=0),reformatFun(dat=demoDatSummary,var="HIVvl",type="medianIQRbignum",rowIdx=2,roundDigits=0))
demoDatKable[demoDatKable$variable=="HIV viral load at 6 months (copies/ml)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="HIVvl_6mts_",type="medianIQRbignum",rowIdx=1,roundDigits=0),reformatFun(dat=demoDatSummary,var="HIVvl_6mts_",type="medianIQRbignum",rowIdx=3,roundDigits=0),reformatFun(dat=demoDatSummary,var="HIVvl_6mts_",type="medianIQRbignum",rowIdx=2,roundDigits=0))
demoDatKable[demoDatKable$variable=="HIV CD4+ at baseline (cells/mm^3^)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="CD4",type="medianIQRbignum",rowIdx=1,roundDigits=0),reformatFun(dat=demoDatSummary,var="CD4",type="medianIQRbignum",rowIdx=3,roundDigits=0),reformatFun(dat=demoDatSummary,var="CD4",type="medianIQRbignum",rowIdx=2,roundDigits=0))
demoDatKable[demoDatKable$variable=="HIV CD4+ at 6 months (cells/mm^3^)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="CD4_6mts_",type="medianIQRbignum",rowIdx=1,roundDigits=0),reformatFun(dat=demoDatSummary,var="CD4_6mts_",type="medianIQRbignum",rowIdx=3,roundDigits=0),reformatFun(dat=demoDatSummary,var="CD4_6mts_",type="medianIQRbignum",rowIdx=2,roundDigits=0))
demoDatKable[demoDatKable$variable=="CD163 (ng/ml)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="CD163",type="medianIQRbignum",rowIdx=1),reformatFun(dat=demoDatSummary,var="CD163",type="medianIQRbignum",rowIdx=3),reformatFun(dat=demoDatSummary,var="CD163",type="medianIQRbignum",rowIdx=2))
demoDatKable[demoDatKable$variable=="ICAM-1 (pg/ml)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="ICAM1",type="medianIQRbignum",rowIdx=1),reformatFun(dat=demoDatSummary,var="ICAM1",type="medianIQRbignum",rowIdx=3),reformatFun(dat=demoDatSummary,var="ICAM1",type="medianIQRbignum",rowIdx=2))
demoDatKable[demoDatKable$variable=="IL-6 (pg/ml)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="IL6",type="medianIQRbignum",rowIdx=1),reformatFun(dat=demoDatSummary,var="IL6",type="medianIQRbignum",rowIdx=3),reformatFun(dat=demoDatSummary,var="IL6",type="medianIQRbignum",rowIdx=2))
demoDatKable[demoDatKable$variable=="IL-1b (pg/ml)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="IL1b",type="medianIQRbignum",rowIdx=1),reformatFun(dat=demoDatSummary,var="IL1b",type="medianIQRbignum",rowIdx=3),reformatFun(dat=demoDatSummary,var="IL1b",type="medianIQRbignum",rowIdx=2))
demoDatKable[demoDatKable$variable=="CMV (AU/ml)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="cmv",type="medianIQR",rowIdx=1),reformatFun(dat=demoDatSummary,var="cmv",type="medianIQR",rowIdx=3),reformatFun(dat=demoDatSummary,var="cmv",type="medianIQR",rowIdx=2))
demoDatKable[demoDatKable$variable=="CMV status",-c(1:2)]<-c(paste(sep="",format(nsmall=1,round(digits=1,100*demoDatSummary[1,"cmvStatusMV"])),"%"),NA,NA,paste(sep="",format(nsmall=1,round(digits=1,100*demoDatSummary[3,"cmvStatusMV"])),"%"),NA,NA,paste(sep="",format(nsmall=1,round(digits=1,100*demoDatSummary[2,"cmvStatusMV"])),"%"),NA,NA)
demoDatKable[demoDatKable$variable=="CMV status: persistently low titres",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="cmvStatusPLow",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="cmvStatusPLow",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="cmvStatusPLow",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="CMV status: progression to low titres",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="cmvStatusPLat",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="cmvStatusPLat",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="cmvStatusPLat",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="CMV status: progression to high titres",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="cmvStatusRLat",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="cmvStatusRLat",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="cmvStatusRLat",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="CMV status: persistently high titres",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="cmvStatusPHigh",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="cmvStatusPHigh",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="cmvStatusPHigh",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="VZV (mIU/ml)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="vzv",type="medianIQR",rowIdx=1),reformatFun(dat=demoDatSummary,var="vzv",type="medianIQR",rowIdx=3),reformatFun(dat=demoDatSummary,var="vzv",type="medianIQR",rowIdx=2))
demoDatKable[demoDatKable$variable=="VZV status",-c(1:2)]<-c(paste(sep="",format(nsmall=1,round(digits=1,100*demoDatSummary[1,"vzvStatusMV"])),"%"),NA,NA,paste(sep="",format(nsmall=1,round(digits=1,100*demoDatSummary[3,"vzvStatusMV"])),"%"),NA,NA,paste(sep="",format(nsmall=1,round(digits=1,100*demoDatSummary[2,"vzvStatusMV"])),"%"),NA,NA)
demoDatKable[demoDatKable$variable=="VZV status: persistently low titres",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="vzvStatusPLow",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="vzvStatusPLow",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="vzvStatusPLow",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="VZV status: progression to low titres",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="vzvStatusPLat",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="vzvStatusPLat",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="vzvStatusPLat",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="VZV status: progression to high titres",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="vzvStatusRLat",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="vzvStatusRLat",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="vzvStatusRLat",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="VZV status: persistently high titres",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="vzvStatusPHigh",type="nProp",rowIdx=1),reformatFun(dat=demoDatSummary,var="vzvStatusPHigh",type="nProp",rowIdx=3),reformatFun(dat=demoDatSummary,var="vzvStatusPHigh",type="nProp",rowIdx=2))
demoDatKable[demoDatKable$variable=="PWV (m/s)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="PWV",type="meanSD",rowIdx=1),reformatFun(dat=demoDatSummary,var="PWV",type="meanSD",rowIdx=3),reformatFun(dat=demoDatSummary,var="PWV",type="meanSD",rowIdx=2))
demoDatKable[demoDatKable$variable=="CIMT (mm)",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="CIMT",type="meanSD",rowIdx=1),reformatFun(dat=demoDatSummary,var="CIMT",type="meanSD",rowIdx=3),reformatFun(dat=demoDatSummary,var="CIMT",type="meanSD",rowIdx=2))
demoDatKable[demoDatKable$variable=="ABPI",-c(1:2)]<-c(reformatFun(dat=demoDatSummary,var="ABPI",type="meanSD",rowIdx=1),reformatFun(dat=demoDatSummary,var="ABPI",type="meanSD",rowIdx=3),reformatFun(dat=demoDatSummary,var="ABPI",type="meanSD",rowIdx=2))
demoDatKable[demoDatKable$variable=="Confirmed clinical CVD/CBD events",-c(1:2)]<-c(NA,demoDatSummary[1,"clinEventsConfirmedCVD"],format(nsmall=1,round(demoDatSummary[1,"clinEventsConfirmedCVDPer1000PYs"],digits=1)),NA,demoDatSummary[3,"clinEventsConfirmedCVD"],format(nsmall=1,round(demoDatSummary[3,"clinEventsConfirmedCVDPer1000PYs"],digits=1)),NA,demoDatSummary[2,"clinEventsConfirmedCVD"],format(nsmall=1,round(demoDatSummary[2,"clinEventsConfirmedCVDPer1000PYs"],digits=1)))
demoDatKable[demoDatKable$variable=="Sub-clinical CIMT events",-c(1:2)]<-c(NA,demoDatSummary[1,"clinEventsCimt"],format(nsmall=1,round(demoDatSummary[1,"clinEventsCimtPer1000PYs"],digits=1)),NA,demoDatSummary[3,"clinEventsCimt"],format(nsmall=1,round(demoDatSummary[3,"clinEventsCimtPer1000PYs"],digits=1)),NA,demoDatSummary[2,"clinEventsCimt"],format(nsmall=1,round(demoDatSummary[2,"clinEventsCimtPer1000PYs"],digits=1)))
demoDatKable[demoDatKable$variable=="Sub-/clinical CVD/CBD events",-c(1:2)]<-c(NA,demoDatSummary[1,"clinEventsConfirmedCVDWithCIMT"],format(nsmall=1,round(demoDatSummary[1,"clinEventsConfirmedCVDWithCIMTPer1000PYs"],digits=1)),NA,demoDatSummary[3,"clinEventsConfirmedCVDWithCIMT"],format(nsmall=1,round(demoDatSummary[3,"clinEventsConfirmedCVDWithCIMTPer1000PYs"],digits=1)),NA,demoDatSummary[2,"clinEventsConfirmedCVDWithCIMT"],format(nsmall=1,round(demoDatSummary[2,"clinEventsConfirmedCVDWithCIMTPer1000PYs"],digits=1)))

options(knitr.kable.NA = '')

demoDatKable %>%
  kable(col.names=rep("",ncol(demoDatKable)),caption="Participant characteristics (at baseline).") %>%
  kable_styling(full_width=F) %>%
  add_header_above(c(" "," ","Missing values"," "=2,"Missing values"," "=2,"Missing values"," "=2)) %>%
  add_header_above(c(" "," ","All"=3,"PLWH"=3,"Controls"=3)) %>%
  pack_rows("Non-modifiable risk factors",3,8) %>%
  pack_rows("Modifiable risk factors",9,23) %>%
  pack_rows("Framingham risk score",24,24) %>%
  pack_rows("Exposure variables",25,29) %>%
  pack_rows("Mediators",30,33) %>%
  pack_rows("Mediators - herpes",34,45) %>%
  pack_rows("Outcomes",46,49)

Table 6.3: Participant characteristics (at baseline).
		All			PLWH			Controls
		Missing values			Missing values			Missing values

Participants	n (% out of total)		896	(100.0%)		710	(79.2%)		186	(20.8%)
Followed-up at 24 months	n (%)		671	(74.9%)		520	(73.2%)		151	(81.2%)
Non-modifiable risk factors
Sex: male	n (%)	0.0%	540	(60.3%)	0.0%	445	(62.7%)	0.0%	95	(51.1%)
Age	median (IQR)	0.0%	42.0	(38.0-48.0)	0.0%	42.0	(38.0-47.0)	0.0%	43.0	(38.2-49.0)
Age group [35,45)	n (%)		556	(62.1%)		449	(63.2%)		107	(57.5%)
Age group [45,55)	n (%)		244	(27.2%)		196	(27.6%)		48	(25.8%)
Age group [55,65)	n (%)		67	(7.5%)		51	(7.2%)		16	(8.6%)
Age group 65+	n (%)		29	(3.2%)		14	(2.0%)		15	(8.1%)
Modifiable risk factors
BMI (kg/m²)	median (IQR)	0.0%	21.7	(19.6-25.2)	0.0%	21.2	(19.4-24.0)	0.0%	25.3	(21.6-29.8)
Waist-hip ratio	median (IQR)	0.0%	0.8	(0.8-0.9)	0.0%	0.8	(0.8-0.9)	0.0%	0.9	(0.8-0.9)
Diastolic blood pressure (mm Hg)	mean (SD)	0.0%	78.5	(12.4)	0.0%	77.4	(12.2)	0.0%	82.9	(12.0)
Systolic blood pressure (mm Hg)	mean (SD)	0.0%	117.9	(20.0)	0.0%	115.6	(19.3)	0.0%	126.8	(20.3)
Blood pressure: normotensive	n (%)		465	(51.9%)		403	(56.8%)		62	(33.3%)
Blood pressure: pre-hypertensive	n (%)		272	(30.4%)		197	(27.7%)		75	(40.3%)
Blood pressure: hypertensive	n (%)		159	(17.7%)		110	(15.5%)		49	(26.3%)
Cholesterol (mmol/l)	median (IQR)	1.3%	3.8	(3.2-4.6)	1.3%	3.6	(3.0-4.3)	1.6%	4.8	(4.0-5.5)
Hypercholesteremic	n (%)		118	(13.3%)		60	(8.6%)		58	(31.7%)
Glucose (mmol/l)	mean (SD)	1.1%	5.2	(1.9)	1.4%	5.2	(1.9)	0.0%	5.1	(1.7)
Diabetic	n (%)		9	(1.0%)		6	(0.9%)		3	(1.6%)
eGFR (ml/min/1.73m²)	mean (SD)	0.4%	104.3	(20.3)	0.6%	104.2	(20.5)	0.0%	104.8	(19.3)
Chronic kidney disease	n (%)		40	(4.5%)		32	(4.5%)		8	(4.3%)
Current smoker	n (%)	0.0%	98	(10.9%)	0.0%	81	(11.4%)	0.0%	17	(9.1%)
Drink alcohol	n (%)	0.0%	180	(20.1%)	0.0%	157	(22.1%)	0.0%	23	(12.4%)
Framingham risk score
Framingham 10-year CVD risk	median (IQR)	12.1%	3.6%	(2.0%-6.1%)	11.4%	3.5%	(2.0%-5.8%)	14.5%	4.0%	(2.3%-7.8%)
Exposure variables
HIV infected	n (%)	0.0%	710	(79.2%)	0.0%	710	(100.0%)	0.0%	0	(0.0%)
HIV viral load at baseline (copies/ml)	median (IQR)	23.0%	55,579	(6,472-225,919)	2.8%	55,579	(6,472-225,919)	100.0%	NA	(NA-NA)
HIV viral load at 6 months (copies/ml)	median (IQR)	32.9%	1	(1-1)	15.5%	1	(1-1)	99.5%	1	(1-1)
HIV CD4+ at baseline (cells/mm³)	median (IQR)	21.8%	192	(107-312)	1.3%	192	(107-312)	100.0%	NA	(NA-NA)
HIV CD4+ at 6 months (cells/mm³)	median (IQR)	32.6%	294	(203-436)	15.1%	293	(202-436)	99.5%	441	(441-441)
Mediators
CD163 (ng/ml)	median (IQR)	19.1%	73.4	(51.3-115.3)	18.3%	76.7	(54.0-129.1)	22.0%	61.1	(45.7-80.3)
ICAM-1 (pg/ml)	median (IQR)	18.9%	69,674.6	(40,453.0-150,446.7)	18.0%	68,619.5	(39,395.7-150,593.5)	22.0%	74,474.0	(43,993.5-150,037.4)
IL-6 (pg/ml)	median (IQR)	18.9%	2.4	(1.2-4.0)	18.0%	2.4	(1.2-4.9)	22.0%	1.8	(0.9-2.4)
IL-1b (pg/ml)	median (IQR)	19.0%	2.4	(0.7-2.4)	18.0%	2.4	(0.7-2.4)	22.6%	2.4	(0.7-2.4)
Mediators - herpes
CMV (AU/ml)	median (IQR)	12.7%	322.4	(180.8-868.7)	13.0%	527.5	(198.0-998.5)	11.8%	193.7	(145.2-345.9)
CMV status		43.2%			42.0%			47.8%
CMV status: persistently low titres	n (%)		336	(66.0%)		244	(59.2%)		92	(94.8%)
CMV status: progression to low titres	n (%)		11	(2.2%)		11	(2.7%)		0	(0.0%)
CMV status: progression to high titres	n (%)		68	(13.4%)		65	(15.8%)		3	(3.1%)
CMV status: persistently high titres	n (%)		94	(18.5%)		92	(22.3%)		2	(2.1%)
VZV (mIU/ml)	median (IQR)	12.8%	1143.0	(731.9-1832.0)	13.1%	1235.0	(759.3-1893.0)	11.8%	896.2	(628.8-1365.5)
VZV status		43.2%			41.8%			48.4%
VZV status: persistently low titres	n (%)		343	(67.4%)		267	(64.6%)		76	(79.2%)
VZV status: progression to low titres	n (%)		59	(11.6%)		51	(12.3%)		8	(8.3%)
VZV status: progression to high titres	n (%)		22	(4.3%)		21	(5.1%)		1	(1.0%)
VZV status: persistently high titres	n (%)		85	(16.7%)		74	(17.9%)		11	(11.5%)
Outcomes
PWV (m/s)	mean (SD)	1.8%	9.0	(1.8)	1.3%	9.0	(1.7)	3.8%	9.2	(2.1)
CIMT (mm)	mean (SD)	0.4%	0.5	(0.1)	0.6%	0.5	(0.1)	0.0%	0.5	(0.1)
ABPI	mean (SD)	1.0%	1.0	(0.1)	1.3%	1.0	(0.1)	0.0%	1.0	(0.1)
Confirmed clinical CVD/CBD events	n (events per 1,000 person-years)		9	5.3		8	6.1		1	2.4
Sub-clinical CIMT events	n (events per 1,000 person-years)		11	6.4		9	6.9		2	4.9
Sub-/clinical CVD/CBD events	n (events per 1,000 person-years)		20	11.7		17	13.0		3	7.3

#write.csv(demoDatKable,row.names=F,file=paste(sep="_",outPrefix,"participantCharacteristicsTable_Table1.csv"))
write_excel_csv(demoDatKable,file=paste(sep="_",outPrefix,"participantCharacteristicsTable_Table1.csv"))

6.1.3.1 CMV and VZV categories

In the analyses below we defined categories of CMV / VZV infection status, based on titres measured at enrolment and 6 month:

Persistently low titres: titres at both enrolment and 6 month in the bottom 3 quartiles of the data
Progression to low titres: titres at enrolment in the top quartile of values and titres at 6 months in bottom 3 quartiles
Progression to high titres: titres at enrolment in the bottom 3 quartiles of values and titres at 6 months in the top quartile
Persistently high titres: titres at both enrolment and 6 month in the top quartile of the data

6.1.4 Multiple imputation

We have used the R package mice to perform the multiple imputation. Specifically, we use univariate models via fully conditional specification (FCS); i.e. variable-by-variable imputation. This provides a statistically principled basis to perform analysis.

Missing data patterns are summarised visually on Figure 6.3.

impVars <- c("pid","crf.ptype","crf.age","crf.gender","crf.height",grep(value=T,colnames(rhiccaDatWide),pattern="physio.bmi_enrolment|physio.bmi_[0-9]+months|physio.waisthip|physio.hr|physio.temp|physio.bp_dia|physio.bp_sys|pwv.pwv_enrolment|pwv.pwv_[0-9]+months|cimt.cimt_enrolment|cimt.cimt_[0-9]+months|abpi.abpi_enrolment|abpi.abpi_[0-9]+months|eGFR|lims.CD4_Absolute|lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment|lims.Glucose_enrolment|lims.Glucose_[0-9]+months|lims.Cholesterol|lum.CD163_|lum.ICAM1_|lum.IL6_|lum.ILb_|iris|cmv.cmv_|vzv.vzv_"))

impVars<-impVars[!grepl(impVars,pattern="_30months") & !grepl(impVars,pattern="_36months")]
impVars<-setdiff(impVars,c("cmv.cmv_12months","cmv.cmv_24months","cmv.cmv_NA","vzv.vzv_12months","vzv.vzv_24months","vzv.vzv_NA"))
impVars<-c(impVars,"cmvStatus","vzvStatus")

impDat<-rhiccaDatWide %>%
  dplyr::select(all_of(impVars))
impDat<-impDat %>%
  filter(!is.na(crf.ptype) & !is.na(crf.age) & !is.na(crf.gender))

mdPattern<-md.pattern(impDat,plot=F)
missValsPerPattern=mdPattern[-nrow(mdPattern),ncol(mdPattern)]
missValsPerVar=mdPattern[nrow(mdPattern),-ncol(mdPattern)]
missValsPerVarProp=format(nsmall=2,round(digits=2,mdPattern[nrow(mdPattern),-ncol(mdPattern)]/nrow(impDat)))
mdPattern<-mdPattern[-nrow(mdPattern),-ncol(mdPattern)]
mdPatternIdx<-order(missValsPerPattern)
mdPattern<-mdPattern[mdPatternIdx,]
missValsPerPattern<-missValsPerPattern[mdPatternIdx]

mdPatternGg<-data.frame(
  x=as.vector(matrix(nrow=nrow(mdPattern),ncol=ncol(mdPattern),rep(1:ncol(mdPattern),nrow(mdPattern)),byrow=T)),
  y=rep(nrow(mdPattern):1,ncol(mdPattern)),
  z=as.vector(mdPattern)
)

g<-mdPatternGg %>%
  as.data.frame() %>%
  ggplot(mapping=aes(x=x, y=y, fill=as.factor(z))) +
  geom_tile(color="darkgrey",show.legend=F) +
  geom_text(data=data.frame(x=1:ncol(mdPattern),y=nrow(mdPattern)+1,label=colnames(mdPattern),z=NA),mapping=aes(x=x,y=y,label=label),size=2,angle=60,hjust=0) +
  geom_text(data=data.frame(x=1:ncol(mdPattern),y=0,label=missValsPerVarProp,z=NA),mapping=aes(x=x,y=y,label=label),size=2,angle=90,hjust=1) +
  geom_text(data=data.frame(x=ncol(mdPattern)+1,y=nrow(mdPattern):1,label=missValsPerPattern,z=NA),mapping=aes(x=x,y=y,label=label),size=1.75,angle=0,hjust=0) +
  geom_text(data=data.frame(x=0,y=nrow(mdPattern):1,label=rownames(mdPattern),z=NA),mapping=aes(x=x,y=y,label=label),size=1.75,angle=0,hjust=1) +
  scale_fill_manual(values=c("salmon","steelblue")) +
  guides() +
  theme_void() +
  theme(plot.margin = unit(c(10, 6, 2, 6), "pt")) +
  labs(caption="Missing value patterns in the data. Each column represents one variable.\nThe top annotation states the variable name.\nThe right-hand number indicates the number of missing variables per pattern.\nThe bottom number indicates the proportion of missing data per variable.\nThe left-hand side number indicates the frequency of each pattern in the data.")
# + coord_fixed()

print(g)

Figure 6.3: Missing data patterns in the RHICCA dataset.

pdf(paste(sep="_",outPrefix,"_missingnessPatterns.pdf"),width=15,height=18)
print(g)
dev.off()

impDat$lims.CD4_Absolute_enrolment[impDat$crf.ptype=="Control"]<-0
impDat$lims.CD4_Absolute_6months[impDat$crf.ptype=="Control"]<-0
impDat$lims.CD4_Absolute_12months[impDat$crf.ptype=="Control"]<-0
impDat$lims.CD4_Absolute_24months[impDat$crf.ptype=="Control"]<-0
impDat$lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment[impDat$crf.ptype=="Control"]<-1

impDatLong<-impDat %>%
  dplyr::select(!any_of(c("pwvChange","cimtChange","abpiChange","egfrChange"))) %>%
  pivot_longer(cols=!c(pid,starts_with("crf"),contains("bmi"),contains("waisthip"),lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment,contains("Glucose"),contains("Cholesterol"),lims.CD4_Absolute_enrolment,iris),names_to=c(".value","visit"),names_pattern="([a-zA-z0-9_\\.]+)_(enrolment|[0-9]+months)") %>%
  mutate(visitNum=case_when(
    visit=="enrolment"~0,
    visit=="6months"~6,
    visit=="12months"~12,
    visit=="18months"~18,
    visit=="24months"~24,
    TRUE~NA_real_
  )) %>%
  filter(!is.na(visitNum))

# change from baseline encoding
impDatLong$lims.CD4_Absolute[impDatLong$visitNum==0]<-0
impDatLong$lims.CD4_Absolute[impDatLong$visitNum==6]<-impDatLong$lims.CD4_Absolute[impDatLong$visitNum==6]-impDatLong$lims.CD4_Absolute_enrolment[impDatLong$visitNum==6]
impDatLong$lims.CD4_Absolute[impDatLong$visitNum==12]<-impDatLong$lims.CD4_Absolute[impDatLong$visitNum==12]-impDatLong$lims.CD4_Absolute_enrolment[impDatLong$visitNum==12]
impDatLong$lims.CD4_Absolute[impDatLong$visitNum==18]<-impDatLong$lims.CD4_Absolute[impDatLong$visitNum==18]-impDatLong$lims.CD4_Absolute_enrolment[impDatLong$visitNum==18]
impDatLong$lims.CD4_Absolute[impDatLong$visitNum==24]<-impDatLong$lims.CD4_Absolute[impDatLong$visitNum==24]-impDatLong$lims.CD4_Absolute_enrolment[impDatLong$visitNum==24]

skip.Imp<-TRUE

if(!skip.Imp){
  seed<-202208
  impDat<-impDat %>%
    mutate(
      pwvChange=NA,
      cimtChange=NA,
      abpiChange=NA,
      egfrChange=NA,
      cmvChange=NA,
      vzvChange=NA,
      lims.CD4_Absolute_changeFromBaselineAt24Months=NA,
      lum.CD163_change=NA,
      lum.IL6_change=NA,
      lum.ICAM1_change=NA
    )
  
  impQuickPred <- quickpred(impDat, mincor=0.1, include=c("crf.ptype","crf.age","crf.gender"),exclude=c("pid","iris"))
  meth<-rep("cart",ncol(impDat))
  names(meth)<-colnames(impDat)
  meth["pid"]<-""
  meth["crf.ptype"]<-""
  meth["crf.age"]<-""
  meth["crf.gender"]<-""
  meth["iris"]<-""
  meth[rowSums(impQuickPred)==0]<-""
  meth["pwvChange"]<-"~ I(pwv.pwv_24months - pwv.pwv_enrolment)"
  meth["cimtChange"]<-"~ I(cimt.cimt_24months - cimt.cimt_enrolment)"
  meth["abpiChange"]<-"~ I(abpi.abpi_24months - abpi.abpi_enrolment)"
  meth["egfrChange"]<-"~ I(eGFR_24months - eGFR_enrolment)"
  meth["cmvChange"]<-"~ I(cmv.cmv_6months - cmv.cmv_enrolment)"
  meth["vzvChange"]<-"~ I(vzv.vzv_6months - vzv.vzv_enrolment)"
  meth["lims.CD4_Absolute_changeFromBaselineAt24Months"]<-"~ I(lims.CD4_Absolute_24months-lims.CD4_Absolute_enrolment)"
  meth["lum.CD163_change"]<-"~ I(lum.CD163_6months-lum.CD163_enrolment)"
  meth["lum.IL6_change"]<-"~ I(lum.IL6_6months-lum.IL6_enrolment)"
  meth["lum.ICAM1_change"]<-"~ I(lum.ICAM1_6months-lum.ICAM1_enrolment)"
  
  cat(file=logFile,append=T,"Starting the imputation.\n")
  
  m<-10
  #rhiccaDatWide_impMice <- parlmice(data=impDat, m=m, predictorMatrix=impQuickPred, maxit=5, printFlag=FALSE, n.core=5, n.imp.core=1, cl.type="FORK",method=meth,cluster.seed=seed)
  rhiccaDatWide_impMice <- mice(data=impDat, m=m, predictorMatrix=impQuickPred, maxit=5, printFlag=FALSE, seed=seed, method=meth) # fast enough to not require parallel computations (which get stuck sometimes)
  
  cat(file=logFile,append=T,"Imputation post-processing and saving.\n")
  
  rhiccaDatLong_impMiceStack<-data.frame(impDatLong,.imp=0)
  
  for(i in 1:m){
    tmp<-mice::complete(rhiccaDatWide_impMice,i) %>% select(!any_of(c("pwvChange","cimtChange","abpiChange","egfrChange","cmvChange","vzvChange","lims.CD4_Absolute_changeFromBaselineAt24Months","lum.CD163_change","lum.IL6_change","lum.ICAM1_change")))
    tmpLong<-pivot_longer(tmp,cols=!c(pid,starts_with("crf"),contains("bmi"),contains("waisthip"),lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment,contains("Glucose"),contains("Cholesterol"),lims.CD4_Absolute_enrolment,iris),names_to=c(".value","visit"),names_pattern="([a-zA-z0-9_\\.]+)_(enrolment|[0-9]+months)")
    tmpLong<-tmpLong %>% mutate(visitNum=case_when(
      visit=="enrolment"~0,
      visit=="6months"~6,
      visit=="12months"~12,
      visit=="18months"~18,
      visit=="24months"~24,
    ))
    
    # change from baseline encoding
    tmpLong$lims.CD4_Absolute[tmpLong$visitNum==0]<-0
    tmpLong$lims.CD4_Absolute[tmpLong$visitNum==6]<-tmpLong$lims.CD4_Absolute[tmpLong$visitNum==6]-tmpLong$lims.CD4_Absolute_enrolment[tmpLong$visitNum==6]
    tmpLong$lims.CD4_Absolute[tmpLong$visitNum==12]<-tmpLong$lims.CD4_Absolute[tmpLong$visitNum==12]-tmpLong$lims.CD4_Absolute_enrolment[tmpLong$visitNum==12]
    tmpLong$lims.CD4_Absolute[tmpLong$visitNum==18]<-tmpLong$lims.CD4_Absolute[tmpLong$visitNum==18]-tmpLong$lims.CD4_Absolute_enrolment[tmpLong$visitNum==18]
    tmpLong$lims.CD4_Absolute[tmpLong$visitNum==24]<-tmpLong$lims.CD4_Absolute[tmpLong$visitNum==24]-tmpLong$lims.CD4_Absolute_enrolment[tmpLong$visitNum==24]
    
    rhiccaDatLong_impMiceStack<-rbind(rhiccaDatLong_impMiceStack,data.frame(tmpLong,.imp=i))
  }
  
  rhiccaDatLong_impMiceStack<-rhiccaDatLong_impMiceStack %>%
    mutate(pidLong=paste(sep="_",pid,visit))
  
  rhiccaDatLong_impMice<-as.mids(rhiccaDatLong_impMiceStack,.imp=".imp",.id="pidLong")
  
  save(list=c("impQuickPred","rhiccaDatWide_impMice","rhiccaDatLong_impMice"),file=paste(sep="",outPrefix,"_multiplyImputedData.RData"))
}else{
  date2use<-"20220811"
  prefixTmp<-paste(sep="","../output/",date2use,"/RHICCA_",date2use)
  load(paste(sep="",prefixTmp,"_multiplyImputedData.RData"))
  
  tmpLong<-mice::complete(rhiccaDatWide_impMice, action='long', include=TRUE) %>%
    mutate(
      recruitDate=as.character(obsTime$recruitDate[match(pid,obsTime$pid)]),
      lastVisitDate=as.character(obsTime$lastVisitDate[match(pid,obsTime$pid)]),
      obsTime=as.numeric(obsTime$obsTime[match(pid,obsTime$pid)]),
      eGFR_existingCKD=factor(rhiccaDatWide$eGFR_existingCKD[match(pid,rhiccaDatWide$pid)]),
      cmvStatus=factor(case_when(
        !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment<cmvQuartiles["75%"] & cmv.cmv_6months<cmvQuartiles["75%"] ~ "persistently low titres",
      !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment>=cmvQuartiles["75%"] & cmv.cmv_6months<cmvQuartiles["75%"] ~ "progression to low titres",
        !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment<cmvQuartiles["75%"] & cmv.cmv_6months>=cmvQuartiles["75%"] ~ "progression to high titres",
        !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment>=cmvQuartiles["75%"] & cmv.cmv_6months>=cmvQuartiles["75%"] ~ "persistently high titres",
        TRUE~NA_character_
      ),levels=c("persistently low titres","progression to low titres","progression to high titres","persistently high titres")),
      vzvStatus=factor(case_when(
        !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment<vzvQuartiles["75%"] & vzv.vzv_6months<vzvQuartiles["75%"] ~ "persistently low titres",
        !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment>=vzvQuartiles["75%"] & vzv.vzv_6months<vzvQuartiles["75%"] ~ "progression to low titres",
        !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment<vzvQuartiles["75%"] & vzv.vzv_6months>=vzvQuartiles["75%"] ~ "progression to high titres",
        !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment>=vzvQuartiles["75%"] & vzv.vzv_6months>=vzvQuartiles["75%"] ~ "persistently high titres",
        TRUE~NA_character_
      ),levels=c("persistently low titres","progression to low titres","progression to high titres","persistently high titres")),
      cmvEnrolmentStatus=factor(case_when(
      !is.na(cmv.cmv_enrolment) & cmv.cmv_enrolment<cmvQuartiles["75%"] ~ "low / medium enrolment titres",
      !is.na(cmv.cmv_enrolment) & cmv.cmv_enrolment>=cmvQuartiles["75%"] ~ "high enrolment titres",
      TRUE~NA_character_
    ),levels=c("low / medium enrolment titres","high enrolment titres")),
      vzvEnrolmentStatus=factor(case_when(
      !is.na(vzv.vzv_enrolment) & vzv.vzv_enrolment<vzvQuartiles["75%"] ~ "low / medium enrolment titres",
      !is.na(vzv.vzv_enrolment) & vzv.vzv_enrolment>=vzvQuartiles["75%"] ~ "high enrolment titres",
      TRUE~NA_character_
    ),levels=c("low / medium enrolment titres","high enrolment titres"))
    ) %>%
  mutate(
    cmvStatus_2levels=factor(case_when(
      !is.na(cmvStatus) & cmvStatus=="persistently high titres"~"persistently high titres",
      !is.na(cmvStatus) & cmvStatus!="persistently high titres"~"not persistently high titres"
    ),levels=c("not persistently high titres","persistently high titres")),
    vzvStatus_2levels=factor(case_when(
      !is.na(vzvStatus) & vzvStatus=="persistently high titres"~"persistently high titres",
      !is.na(vzvStatus) & vzvStatus!="persistently high titres"~"not persistently high titres"
    ),levels=c("not persistently high titres","persistently high titres"))
  )
  
  rhiccaDatWide_impMice <- as.mids(tmpLong)
}

tmpLong<-mice::complete(rhiccaDatLong_impMice, action='long', include=TRUE) %>%
  mutate(
    lum.CD163_enrolment=tmpLong$lum.CD163_enrolment[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    lum.CD163_6months=tmpLong$lum.CD163_6months[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    lum.CD163_change=tmpLong$lum.CD163_6months[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))]-tmpLong$lum.CD163_enrolment[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    lum.IL6_enrolment=tmpLong$lum.IL6_enrolment[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    lum.IL6_6months=tmpLong$lum.IL6_6months[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    lum.IL6_change=tmpLong$lum.IL6_6months[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))]-tmpLong$lum.IL6_enrolment[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    lum.ICAM1_enrolment=tmpLong$lum.ICAM1_enrolment[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    lum.ICAM1_6months=tmpLong$lum.ICAM1_6months[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    lum.ICAM1_change=tmpLong$lum.ICAM1_6months[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))]-tmpLong$lum.ICAM1_enrolment[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    cmv.cmv_enrolment=tmpLong$cmv.cmv_enrolment[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    cmv.cmv_6months=tmpLong$cmv.cmv_6months[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    cmvChange=tmpLong$cmvChange[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    vzv.vzv_enrolment=tmpLong$vzv.vzv_enrolment[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    vzv.vzv_6months=tmpLong$vzv.vzv_6months[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    vzvChange=tmpLong$vzvChange[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    lims.Glucose_enrolment=tmpLong$lims.Glucose_enrolment[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    lims.Cholesterol_enrolment=tmpLong$lims.Cholesterol_enrolment[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))],
    physio.bmi_enrolment=tmpLong$physio.bmi_enrolment[match(paste(sep="_",pid,.imp),paste(sep="_",tmpLong$pid,tmpLong$.imp))]
  ) %>%
  mutate(
    crf.smoke=rhiccaDatWide$crf.smoke[match(pid,rhiccaDatWide$pid)],
    crf.alcohol=rhiccaDatWide$crf.alcohol[match(pid,rhiccaDatWide$pid)],
    eGFR_existingCKD=rhiccaDatWide$eGFR_existingCKD[match(pid,rhiccaDatWide$pid)]
  ) %>%
  mutate(
    cmvStatus=factor(case_when(
      !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment<cmvQuartiles["75%"] & cmv.cmv_6months<cmvQuartiles["75%"] ~ "persistently low titres",
      !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment>=cmvQuartiles["75%"] & cmv.cmv_6months<cmvQuartiles["75%"] ~ "progression to low titres",
      !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment<cmvQuartiles["75%"] & cmv.cmv_6months>=cmvQuartiles["75%"] ~ "progression to high titres",
      !is.na(cmv.cmv_enrolment) & !is.na(cmv.cmv_6months) & cmv.cmv_enrolment>=cmvQuartiles["75%"] & cmv.cmv_6months>=cmvQuartiles["75%"] ~ "persistently high titres",
      TRUE~NA_character_
    ),levels=c("persistently low titres","progression to low titres","progression to high titres","persistently high titres")),
    vzvStatus=factor(case_when(
      !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment<vzvQuartiles["75%"] & vzv.vzv_6months<vzvQuartiles["75%"] ~ "persistently low titres",
      !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment>=vzvQuartiles["75%"] & vzv.vzv_6months<vzvQuartiles["75%"] ~ "progression to low titres",
      !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment<vzvQuartiles["75%"] & vzv.vzv_6months>=vzvQuartiles["75%"] ~ "progression to high titres",
      !is.na(vzv.vzv_enrolment) & !is.na(vzv.vzv_6months) & vzv.vzv_enrolment>=vzvQuartiles["75%"] & vzv.vzv_6months>=vzvQuartiles["75%"] ~ "persistently high titres",
      TRUE~NA_character_
    ),levels=c("persistently low titres","progression to low titres","progression to high titres","persistently high titres")),
    cmvEnrolmentStatus=factor(case_when(
      !is.na(cmv.cmv_enrolment) & cmv.cmv_enrolment<cmvQuartiles["75%"] ~ "low / medium enrolment titres",
      !is.na(cmv.cmv_enrolment) & cmv.cmv_enrolment>=cmvQuartiles["75%"] ~ "high enrolment titres",
      TRUE~NA_character_
    ),levels=c("low / medium enrolment titres","high enrolment titres")),
      vzvEnrolmentStatus=factor(case_when(
      !is.na(vzv.vzv_enrolment) & vzv.vzv_enrolment<vzvQuartiles["75%"] ~ "low / medium enrolment titres",
      !is.na(vzv.vzv_enrolment) & vzv.vzv_enrolment>=vzvQuartiles["75%"] ~ "high enrolment titres",
      TRUE~NA_character_
    ),levels=c("low / medium enrolment titres","high enrolment titres"))
  ) %>%
  mutate(
    cmvStatus_2levels=factor(case_when(
      !is.na(cmvStatus) & cmvStatus=="persistently high titres"~"persistently high titres",
      !is.na(cmvStatus) & cmvStatus!="persistently high titres"~"not persistently high titres"
    ),levels=c("not persistently high titres","persistently high titres")),
    vzvStatus_2levels=factor(case_when(
      !is.na(vzvStatus) & vzvStatus=="persistently high titres"~"persistently high titres",
      !is.na(vzvStatus) & vzvStatus!="persistently high titres"~"not persistently high titres"
    ),levels=c("not persistently high titres","persistently high titres"))
  )

tmpLong<-tmpLong %>%
  dplyr::mutate(crf.artRegimen.2levels=art$artSimple2Level[match(tmpLong$pid,art$pid)])

rhiccaDatLong_impMice <- as.mids(tmpLong)

tmpLong<-complete(rhiccaDatWide_impMice,action="long",include=TRUE)
tmpLong<-tmpLong %>%
  dplyr::mutate(crf.artRegimen.2levels=art$artSimple2Level[match(tmpLong$pid,art$pid)]) %>%
  dplyr::mutate(
    crf.smoke=rhiccaDatWide$crf.smoke[match(pid,rhiccaDatWide$pid)],
    crf.alcohol=rhiccaDatWide$crf.alcohol[match(pid,rhiccaDatWide$pid)]
  )
rhiccaDatWide_impMice<-as.mids(tmpLong)

7 Baseline population have a low cardiovascular risk profile.

7.0.1 Modifiable CVD risk factors and inflammatory and viral markers over time

We can summarise modifiable CVD risk factors over time by plotting their sample means and 95% confidence intervals over time and by participant type, see Figure 7.1. This shows a low cardiovascular risk for both PLWH and controls at baseline.

# bmi, waist-to-hip, diastolic bp, systolic bp, cholesterol, glucose, eGFR
datTmp<-rhiccaDatLong %>%
  dplyr::select(visitNum,crf.ptype,crf.PLWH,physio.bmi,physio.waisthip,physio.bp_dia,physio.bp_sys,lims.Cholesterol,lims.Glucose,eGFR,cmv.cmv,vzv.vzv,lum.CD163,lum.ICAM1,lum.IL6,lims.CD4_Absolute,lims.HIV_Abbott_Viral_Load_0pt2ml_P) %>%
  filter(!is.na(crf.ptype) & (is.na(lims.Cholesterol) | lims.Cholesterol<2000)) %>%
  group_by(crf.PLWH,visitNum) %>%
  summarise(bmiN=sum(!is.na(physio.bmi)),
            bmiMean=mean(physio.bmi,na.rm=T),
            bmiSD=sd(physio.bmi,na.rm=T),
            bmiMedian=median(physio.bmi,na.rm=T),
            bmi25=quantile(physio.bmi,probs=0.25,na.rm=T),
            bmi75=quantile(physio.bmi,probs=0.75,na.rm=T),
            whrN=sum(!is.na(physio.waisthip)),
            whrMean=mean(physio.waisthip,na.rm=T),
            whrSD=sd(physio.waisthip,na.rm=T),
            whrMedian=median(physio.waisthip,na.rm=T),
            whr25=quantile(physio.waisthip,probs=0.25,na.rm=T),
            whr75=quantile(physio.waisthip,probs=0.75,na.rm=T),
            bpDiaN=sum(!is.na(physio.bp_dia)),
            bpDiaMean=mean(physio.bp_dia,na.rm=T),
            bpDiaSD=sd(physio.bp_dia,na.rm=T),
            bpDiaMedian=median(physio.bp_dia,na.rm=T),
            bpDia25=quantile(physio.bp_dia,probs=0.25,na.rm=T),
            bpDia75=quantile(physio.bp_dia,probs=0.75,na.rm=T),
            bpSysN=sum(!is.na(physio.bp_sys)),
            bpSysMean=mean(physio.bp_sys,na.rm=T),
            bpSysSD=sd(physio.bp_sys,na.rm=T),
            bpSysMedian=median(physio.bp_sys,na.rm=T),
            bpSys25=quantile(physio.bp_sys,probs=0.25,na.rm=T),
            bpSys75=quantile(physio.bp_sys,probs=0.75,na.rm=T),
            cholN=sum(!is.na(lims.Cholesterol)),
            cholMean=mean(lims.Cholesterol,na.rm=T),
            cholSD=sd(lims.Cholesterol,na.rm=T),
            cholMedian=median(lims.Cholesterol,na.rm=T),
            chol25=quantile(lims.Cholesterol,probs=0.25,na.rm=T),
            chol75=quantile(lims.Cholesterol,probs=0.75,na.rm=T),
            glucN=sum(!is.na(lims.Glucose)),
            glucMean=mean(lims.Glucose,na.rm=T),
            glucSD=sd(lims.Glucose,na.rm=T),
            glucMedian=median(lims.Glucose,na.rm=T),
            gluc25=quantile(lims.Glucose,probs=0.25,na.rm=T),
            gluc75=quantile(lims.Glucose,probs=0.75,na.rm=T),
            egfrN=sum(!is.na(eGFR)),
            egfrMean=mean(eGFR,na.rm=T),
            egfrSD=sd(eGFR,na.rm=T),
            egfrMedian=median(eGFR,na.rm=T),
            egfr25=quantile(eGFR,probs=0.25,na.rm=T),
            egfr75=quantile(eGFR,probs=0.75,na.rm=T),
            cmvN=sum(!is.na(cmv.cmv)),
            cmvMean=mean(cmv.cmv,na.rm=T),
            cmvSD=sd(cmv.cmv,na.rm=T),
            cmvMedian=median(cmv.cmv,na.rm=T),
            cmv25=quantile(cmv.cmv,probs=0.25,na.rm=T),
            cmv75=quantile(cmv.cmv,probs=0.75,na.rm=T),
            vzvN=sum(!is.na(vzv.vzv)),
            vzvMean=mean(vzv.vzv,na.rm=T),
            vzvSD=sd(vzv.vzv,na.rm=T),
            vzvMedian=median(vzv.vzv,na.rm=T),
            vzv25=quantile(vzv.vzv,probs=0.25,na.rm=T),
            vzv75=quantile(vzv.vzv,probs=0.75,na.rm=T),
            cd163N=sum(!is.na(lum.CD163)),
            cd163Mean=mean(lum.CD163,na.rm=T),
            cd163SD=sd(lum.CD163,na.rm=T),
            cd163Median=median(lum.CD163,na.rm=T),
            cd16325=quantile(lum.CD163,probs=0.25,na.rm=T),
            cd16375=quantile(lum.CD163,probs=0.75,na.rm=T),
            icam1N=sum(!is.na(lum.ICAM1)),
            icam1Mean=mean(lum.ICAM1,na.rm=T),
            icam1SD=sd(lum.ICAM1,na.rm=T),
            icam1Median=median(lum.ICAM1,na.rm=T),
            icam125=quantile(lum.ICAM1,probs=0.25,na.rm=T),
            icam175=quantile(lum.ICAM1,probs=0.75,na.rm=T),
            il6N=sum(!is.na(lum.IL6)),
            il6Mean=mean(lum.IL6,na.rm=T),
            il6SD=sd(lum.IL6,na.rm=T),
            il6Median=median(lum.IL6,na.rm=T),
            il625=quantile(lum.IL6,probs=0.25,na.rm=T),
            il675=quantile(lum.IL6,probs=0.75,na.rm=T),
            cd4N=sum(!is.na(lims.CD4_Absolute)),
            cd4Mean=mean(lims.CD4_Absolute,na.rm=T),
            cd4SD=sd(lims.CD4_Absolute,na.rm=T),
            cd4Median=median(lims.CD4_Absolute,na.rm=T),
            cd425=quantile(lims.CD4_Absolute,probs=0.25,na.rm=T),
            cd475=quantile(lims.CD4_Absolute,probs=0.75,na.rm=T),
            vlN=sum(!is.na(lims.HIV_Abbott_Viral_Load_0pt2ml_P)),
            vlMean=mean(lims.HIV_Abbott_Viral_Load_0pt2ml_P,na.rm=T),
            vlSD=sd(lims.HIV_Abbott_Viral_Load_0pt2ml_P,na.rm=T),
            vlMedian=median(lims.HIV_Abbott_Viral_Load_0pt2ml_P,na.rm=T),
            vl25=quantile(lims.HIV_Abbott_Viral_Load_0pt2ml_P,probs=0.25,na.rm=T),
            vl75=quantile(lims.HIV_Abbott_Viral_Load_0pt2ml_P,probs=0.75,na.rm=T),
            .groups="drop") %>%
  pivot_longer(cols=c(bmiN,bmiMean,bmiSD,bmiMedian,bmi25,bmi75,
                      whrN,whrMean,whrSD,whrMedian,whr25,whr75,
                      bpDiaN,bpDiaMean,bpDiaSD,bpDiaMedian,bpDia25,bpDia75,
                      bpSysN,bpSysMean,bpSysSD,bpSysMedian,bpSys25,bpSys75,
                      cholN,cholMean,cholSD,cholMedian,chol25,chol75,
                      glucN,glucMean,glucSD,glucMedian,gluc25,gluc75,
                      egfrN,egfrMean,egfrSD,egfrMedian,egfr25,egfr75,
                      cmvN,cmvMean,cmvSD,cmvMedian,cmv25,cmv75,
                      vzvN,vzvMean,vzvSD,vzvMedian,vzv25,vzv75,
                      cd163N,cd163Mean,cd163SD,cd163Median,cd16325,cd16375,
                      icam1N,icam1Mean,icam1SD,icam1Median,icam125,icam175,
                      il6N,il6Mean,il6SD,il6Median,il625,il675,
                      cd4N,cd4Mean,cd4SD,cd4Median,cd425,cd475,
                      vlN,vlMean,vlSD,vlMedian,vl25,vl75),
               names_to="var",
               values_to="value")

datTmpN<-datTmp %>%
  filter(grepl(var,pattern="N")) %>%
  mutate(var=gsub(var,pattern="N",replacement=""))

datTmpMean<-datTmp %>%
  filter(grepl(var,pattern="Mean")) %>%
  mutate(var=gsub(var,pattern="Mean",replacement=""))

datTmpSd<-datTmp %>%
  filter(grepl(var,pattern="SD")) %>%
  mutate(var=gsub(var,pattern="SD",replacement=""))

datTmpMedian<-datTmp %>%
  filter(grepl(var,pattern="Median")) %>%
  mutate(var=gsub(var,pattern="Median",replacement=""))

datTmp25<-datTmp %>%
  filter(grepl(var,pattern="25")) %>%
  mutate(var=gsub(var,pattern="25",replacement=""))

datTmp75<-datTmp %>%
  filter(grepl(var,pattern="75")) %>%
  mutate(var=gsub(var,pattern="75",replacement=""))

datTmp<-left_join(by=c("crf.PLWH","visitNum","var"),datTmpN,datTmpMean,suffix=c(".n",".mean"))
datTmp<-left_join(by=c("crf.PLWH","visitNum","var"),datTmp,datTmpSd,suffix=c("",".sd"))
datTmp<-left_join(by=c("crf.PLWH","visitNum","var"),datTmp,datTmpMedian,suffix=c("",".median"))
datTmp<-left_join(by=c("crf.PLWH","visitNum","var"),datTmp,datTmp25,suffix=c("",".25"))
datTmp<-left_join(by=c("crf.PLWH","visitNum","var"),datTmp,datTmp75,suffix=c("",".75"))

datTmp<-datTmp %>%
  mutate(var=case_when(
    var=="bmi"~"BMI",
    var=="bpDia"~"Bp diastolic",
    var=="bpSys"~"Bp systolic",
    var=="chol"~"Cholesterol",
    var=="egfr"~"eGFR",
    var=="gluc"~"Glucose",
    var=="whr"~"Waist-to-hip ratio",
    var=="cmv"~"CMV",
    var=="vzv"~"VZV",
    var=="cd163"~"CD163",
    var=="icam1"~"ICAM-1",
    var=="il6"~"IL-6",
    var=="cd4"~"CD4+",
    var=="vl"~"HIV viral load",
  ))

colnames(datTmp)<-c("crf.PLWH","visitNum","var","n","mean","sd","median","q25","q75")

datTmp[datTmp$var=="CD4+" & datTmp$crf.PLWH=="Control",4]<-0
datTmp[datTmp$var=="CD4+" & datTmp$crf.PLWH=="Control",5:9]<-NA
datTmp[datTmp$var=="HIV viral load" & datTmp$crf.PLWH=="Control",4]<-0
datTmp[datTmp$var=="HIV viral load" & datTmp$crf.PLWH=="Control",5:9]<-NA

facet_labels <- c("A.  CD4+","B.  HIV viral load","C.  BMI","D.  CMV","E.  Waist-to-hip ratio","F.  VZV","G.  Bp diastolic","H.  CD163","I.  Bp systolic","J.  ICAM-1","K.  Cholesterol","L.  IL-6","M.  Glucose")
names(facet_labels) <- c("CD4+","HIV viral load","BMI","CMV","Waist-to-hip ratio","VZV","Bp diastolic","CD163","Bp systolic","ICAM-1","Cholesterol","IL-6","Glucose")

datTmp$var<-factor(datTmp$var,levels=c("CD4+","HIV viral load","BMI","CMV","Waist-to-hip ratio","VZV","Bp diastolic","CD163","Bp systolic","ICAM-1","Cholesterol","IL-6","Glucose"))

ggModRiskFactors<-datTmp %>%
  #dplyr::filter(var!="CD4+" & var!="HIV viral load") %>%
  dplyr::filter(var!="eGFR" & n!=0 & visitNum<=24) %>%
  dplyr::filter(!(var=="CMV" & visitNum>6)) %>%
  dplyr::filter(!(var=="VZV" & visitNum>6)) %>%
  ggplot(mapping=aes(x=visitNum,y=mean,ymin=mean-1.96*sd/sqrt(n),ymax=mean+1.96*sd/sqrt(n),col=crf.PLWH,fill=crf.PLWH)) +
  #ggplot(mapping=aes(x=visitNum,y=median,ymin=q25,ymax=q75,col=crf.PLWH,fill=crf.PLWH)) +
  geom_point(size=2) +
  geom_line(lwd=1,na.rm=T) +
  geom_ribbon(alpha=0.25,linetype=0,na.rm=T) +
  scale_color_manual(values=c("steelblue","orange"),name="") +
  scale_fill_manual(values=c("steelblue","orange"),name="") +
  guides(fill="none",lty="none",pch="none") +
  theme(
    text=element_text(size=16),
    legend.position = "bottom",
    strip.text.x = element_text(hjust = 0, margin=margin(l=0))
  ) +
  xlab("Study visit (months since enrolment)") +
  ylab("") +
  xlim(c(0,24)) +
  facet_wrap(~var,labeller = labeller(var = facet_labels),scales="free_y",nrow=7,ncol=2)

print(ggModRiskFactors)

Figure 7.1: Modifiable CVD risk factors over time and by participant type.

pdf(width=12,height=12,file=paste(sep="_",outPrefix,"CIMT_CVDevents.pdf"))
print(ggModRiskFactors)
dev.off()

## quartz_off_screen 
##                 2

In PLWH we see appropriate viral suppression and immune restoration within 6 months of initiating ART (top 2 panels of Figure 7.1. Therefore, it raises the question of other factors potentially driving CVD events.

8 Risk of asymptomatic CVD: surrogate markers CIMT and PWV

We can summarise both CIMT and PWV progression over the study period in both PLWH and controls and we can fit regression models to interrogate associations with HIV status and other CVD risk factors.

A key finding, for asymptomatic CVD, we saw a bigger increased risk over time with CIMT in PLWH compared to controls (Table 8.1) but did not observe a similar trend for PWV (Table 8.9). We note that this association disappears in the adjusted model.

We also saw a trend toward associations with CMV reactivation and persistently elevated VZV (pVZV) in unadjusted models , Table 8.1, raising the question whether pVZV may play a role. In the adjusted models for PLWH only, the pVZV association becomes statistically significant.

8.1 Carotid intima-media thickness (CIMT)

Our data show that CIMT on average increased between baseline and 24 months. Further, our data show that there is evidence for a weak association between participant type (PLWH or control) and change-from-baseline CIMT with PLWH having on average a larger increase in CIMT over the study period. This association is lost after adjusting for cardiovascular risk factors.

Below we tabulate the average relative increase in CIMT from baseline to 24 month by participant type.

rhiccaDatWide %>% group_by(crf.ptype) %>%
  summarise(baselineCIMT=mean(na.rm=TRUE,(cimt.cimt_24months-cimt.cimt_enrolment)/cimt.cimt_enrolment),.groups="drop")

## # A tibble: 2 × 2
##   crf.ptype baselineCIMT
##   <fct>            <dbl>
## 1 Control         0.0348
## 2 Patient         0.0645

8.1.1 Descriptive summaries

Figures 8.1, 8.2, 8.3, 8.4 and 8.5 summarise the distribution of CIMT between participant types at the different timepoints. Note that the large CIMT values (~2.0) are observed at unscheduled visits for two specific participants: RE10453, RE1096I.

tmpDatAll <- rhiccaDatLong %>%
  dplyr::select(pid,crf.PLWH,physio.bmi_cat_enrolment,physio.hypertension_cat_enrolment,visit,pwv.pwv,cimt.cimt,abpi.abpi,eGFR,lims.CD4_Absolute) %>%
  filter(!is.na(crf.PLWH) & is.element(visit,set=c("enrolment","6months","12months","18months","24months","30months","36months"))) %>%
  group_by(visit,crf.PLWH) %>%
  summarise(
    pwvMean=mean(pwv.pwv,na.rm=T),pwvSE=sd(pwv.pwv,na.rm=T)/sqrt(sum(!is.na(pwv.pwv))),
    cimtMean=mean(cimt.cimt,na.rm=T),cimtSE=sd(cimt.cimt,na.rm=T)/sqrt(sum(!is.na(cimt.cimt))),
    abpiMean=mean(abpi.abpi,na.rm=T),abpiSE=sd(abpi.abpi,na.rm=T)/sqrt(sum(!is.na(abpi.abpi))),
    egfrMean=mean(eGFR,na.rm=T),egfrSE=sd(eGFR,na.rm=T)/sqrt(sum(!is.na(eGFR))),
    cd4Mean=mean(lims.CD4_Absolute,na.rm=T),cd4SE=sd(lims.CD4_Absolute,na.rm=T)/sqrt(sum(!is.na(lims.CD4_Absolute))),
            .groups="drop") %>%
  filter(visit!="unscheduled") %>%
  mutate(visitNum=ifelse(visit=="enrolment",0,as.integer(gsub(pattern="months",replacement="",visit))))

tmpDatBmi <- rhiccaDatLong %>%
  dplyr::select(pid,crf.PLWH,physio.bmi_cat_enrolment,physio.hypertension_cat_enrolment,visit,pwv.pwv,cimt.cimt,abpi.abpi,eGFR,lims.CD4_Absolute) %>%
  filter(!is.na(crf.PLWH) & is.element(visit,set=c("enrolment","6months","12months","18months","24months","30months","36months"))) %>%
  group_by(visit,crf.PLWH,physio.bmi_cat_enrolment) %>%
  summarise(
    pwvMean=mean(pwv.pwv,na.rm=T),pwvSE=sd(pwv.pwv,na.rm=T)/sqrt(sum(!is.na(pwv.pwv))),
    cimtMean=mean(cimt.cimt,na.rm=T),cimtSE=sd(cimt.cimt,na.rm=T)/sqrt(sum(!is.na(cimt.cimt))),
    abpiMean=mean(abpi.abpi,na.rm=T),abpiSE=sd(abpi.abpi,na.rm=T)/sqrt(sum(!is.na(abpi.abpi))),
    egfrMean=mean(eGFR,na.rm=T),egfrSE=sd(eGFR,na.rm=T)/sqrt(sum(!is.na(eGFR))),
    cd4Mean=mean(lims.CD4_Absolute,na.rm=T),cd4SE=sd(lims.CD4_Absolute,na.rm=T)/sqrt(sum(!is.na(lims.CD4_Absolute))),
            .groups="drop") %>%
  filter(visit!="unscheduled") %>%
  mutate(visitNum=ifelse(visit=="enrolment",0,as.integer(gsub(pattern="months",replacement="",visit))))

tmpDatHyper <- rhiccaDatLong %>%
  dplyr::select(pid,crf.PLWH,physio.bmi_cat_enrolment,physio.hypertension_cat_enrolment,visit,pwv.pwv,cimt.cimt,abpi.abpi,eGFR,lims.CD4_Absolute) %>%
  filter(!is.na(crf.PLWH) & is.element(visit,set=c("enrolment","6months","12months","18months","24months","30months","36months"))) %>%
  group_by(visit,crf.PLWH,physio.hypertension_cat_enrolment) %>%
  summarise(
    pwvMean=mean(pwv.pwv,na.rm=T),pwvSE=sd(pwv.pwv,na.rm=T)/sqrt(sum(!is.na(pwv.pwv))),
    cimtMean=mean(cimt.cimt,na.rm=T),cimtSE=sd(cimt.cimt,na.rm=T)/sqrt(sum(!is.na(cimt.cimt))),
    abpiMean=mean(abpi.abpi,na.rm=T),abpiSE=sd(abpi.abpi,na.rm=T)/sqrt(sum(!is.na(abpi.abpi))),
    egfrMean=mean(eGFR,na.rm=T),egfrSE=sd(eGFR,na.rm=T)/sqrt(sum(!is.na(eGFR))),
    cd4Mean=mean(lims.CD4_Absolute,na.rm=T),cd4SE=sd(lims.CD4_Absolute,na.rm=T)/sqrt(sum(!is.na(lims.CD4_Absolute))),
            .groups="drop") %>%
  filter(visit!="unscheduled") %>%
  mutate(visitNum=ifelse(visit=="enrolment",0,as.integer(gsub(pattern="months",replacement="",visit))))

rhiccaDatLong %>%
  as.data.frame() %>%
  filter(!is.na(crf.PLWH)) %>%
  histFun(var="cimt.cimt",groupVar="crf.PLWH",yLab="CIMT (mm)",labels=c("Control","PLWH"),title="CIMT (all visits)")

Figure 8.1: Histogram of CIMT values across all visits, overall and by participant type.

g1<-rhiccaDatWide %>%
  as.data.frame() %>%
  filter(!is.na(crf.ptype)) %>%
  violinPlotFun(var="cimt.cimt_enrolment",groupVar="crf.PLWH",yLab="CIMT (mm)",labels=c("Control","PLWH"),ylim=c(0.25,1.15),title="CIMT at baseline")

g24<-rhiccaDatWide %>%
  as.data.frame() %>%
  filter(!is.na(crf.ptype)) %>%
  violinPlotFun(var="cimt.cimt_24months",groupVar="crf.PLWH",yLab="CIMT (mm)",labels=c("Control","PLWH"),ylim=c(0.25,1.15),title="CIMT at 24 months")

grid.arrange(g1,g24,nrow=1)

Figure 8.2: Violin plots of CIMT measurements at baseline and 24 months.

tmpDatAllAdj<-rhiccaDatLong %>%
  filter(visitNum==0 | visitNum==24) %>%
  dplyr::select(pid,visitNum,crf.PLWH,pwv.pwv,cimt.cimt,physio.bmi_cat_enrolment,physio.hypertension_cat_enrolment) %>%
  mutate(
    physio.bmi_enrolment=rhiccaDatWide$physio.bmi_enrolment[match(pid,rhiccaDatWide$pid)],
    physio.bp_dia_enrolment=rhiccaDatWide$physio.bp_dia_enrolment[match(pid,rhiccaDatWide$pid)],
    physio.bp_sys_enrolment=rhiccaDatWide$physio.bp_sys_enrolment[match(pid,rhiccaDatWide$pid)],
    lims.Glucose_enrolment=rhiccaDatWide$lims.Glucose_enrolment[match(pid,rhiccaDatWide$pid)],
    lims.Cholesterol_enrolment=rhiccaDatWide$lims.Cholesterol_enrolment[match(pid,rhiccaDatWide$pid)]
  ) %>%
  filter(rowSums(is.na(.))==0)
tmpDatAllAdj<-tmpDatAllAdj[order(tmpDatAllAdj$pid),]

geeModCimt<-geeglm(cimt.cimt~physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment,data=tmpDatAllAdj,id=factor(tmpDatAllAdj$pid),corstr="exchangeable")

tmpDatAllAdj<-tmpDatAllAdj %>%
  mutate(
    cimt.geeResid=residuals(geeModCimt)
  ) %>%
  group_by(visitNum,crf.PLWH) %>%
  summarise(pwvMean=mean(pwv.pwv,na.rm=T),pwvSE=sd(pwv.pwv,na.rm=T)/sqrt(sum(!is.na(pwv.pwv))),
            cimtMean=mean(cimt.cimt,na.rm=T),cimtSE=sd(cimt.cimt,na.rm=T)/sqrt(sum(!is.na(cimt.cimt))),
            cimtAdjMean=mean(cimt.geeResid,na.rm=T),cimtAdjSE=sd(cimt.geeResid,na.rm=T)/sqrt(sum(!is.na(cimt.geeResid))),
            .groups="drop") 

gCIMT<-tmpDatAll %>%
  filter(visitNum==0 | visitNum==24) %>%
  ggplot(mapping=aes(x=visitNum,y=cimtMean,ymin=cimtMean-1.96*cimtSE,ymax=cimtMean+1.96*cimtSE,col=crf.PLWH,pch=crf.PLWH,lty=crf.PLWH,fill=crf.PLWH)) +
  geom_point(size=3) +
  geom_line(lwd=1) +
  geom_ribbon(alpha=0.25,linetype=0) +
  scale_color_manual(values=c("steelblue","orange"),name="HIV status") +
  scale_fill_manual(values=c("steelblue","orange"),name="HIV status") +
  guides(fill="none",lty="none",pch="none") +
  ggtitle("CIMT mean and 95% CI over time by participant group.") + 
  theme(text=element_text(size=16)) +
  xlab("Study visit (months since enrolment)") +
  ylab("CIMT (mm)") +
  xlim(c(0,24))

gCIMTAdj<-tmpDatAllAdj %>%
  filter(visitNum==0 | visitNum==24) %>%
  ggplot(mapping=aes(x=visitNum,y=cimtAdjMean,ymin=cimtAdjMean-1.96*cimtAdjSE,ymax=cimtAdjMean+1.96*cimtAdjSE,col=crf.PLWH,pch=crf.PLWH,lty=crf.PLWH,fill=crf.PLWH)) +
  geom_point(size=3) +
  geom_line(lwd=1) +
  geom_ribbon(alpha=0.25,linetype=0) +
  scale_color_manual(values=c("steelblue","orange"),name="HIV status") +
  scale_fill_manual(values=c("steelblue","orange"),name="HIV status") +
  guides(fill="none",lty="none",pch="none") +
  ggtitle("Adjusted CIMT mean and 95% CI over time by participant group.") + 
  theme(text=element_text(size=16)) +
  xlab("Study visit (months since enrolment)") +
  ylab("Residual CIMT (mm)") +
  labs(caption="Longitudinal CIMT has been adjusted for baseline CVD risk factors\n(waist-hip ratio, systolic and diastolic blood pressure, glucose, cholesterol)\nusing a GEE model with exchangeable correlation structure.") +
  xlim(c(0,24))
  
gCIMTAdj.notitle<-tmpDatAllAdj %>%
  filter(visitNum==0 | visitNum==24) %>%
  ggplot(mapping=aes(x=visitNum,y=cimtAdjMean,ymin=cimtAdjMean-1.96*cimtAdjSE,ymax=cimtAdjMean+1.96*cimtAdjSE,col=crf.PLWH,pch=crf.PLWH,lty=crf.PLWH,fill=crf.PLWH)) +
  geom_point(size=3) +
  geom_line(lwd=1) +
  geom_ribbon(alpha=0.25,linetype=0) +
  scale_color_manual(values=c("steelblue","orange"),name="HIV status") +
  scale_fill_manual(values=c("steelblue","orange"),name="HIV status") +
  guides(fill="none",lty="none",pch="none",col="none") +
  theme(text=element_text(size=12),legend.position="left") +
  xlab("Study visit (months since enrolment)") +
  ylab("Residual CIMT (mm)") +
  #labs(caption="Longitudinal CIMT has been adjusted for baseline CVD risk factors\n(waist-hip ratio, systolic and diastolic blood pressure, glucose, cholesterol)\nusing a GEE model with exchangeable correlation structure.") +
  xlim(c(0,24))
  
grid.arrange(gCIMT,gCIMTAdj,nrow=1)

Figure 8.3: Raw and adjusted CIMT values observed over time, summarised as sample means and 95% confidence levels.

gCIMTbmi<-tmpDatBmi %>%
  filter(visitNum==0 | visitNum==24) %>%
  ggplot(mapping=aes(x=visitNum,y=cimtMean,ymin=cimtMean-1.96*cimtSE,ymax=cimtMean+1.96*cimtSE,col=crf.PLWH,pch=crf.PLWH,lty=crf.PLWH,fill=crf.PLWH)) +
  geom_point(size=3) +
  geom_line(lwd=1) +
  geom_ribbon(alpha=0.25,linetype=0) +
  scale_color_manual(values=c("steelblue","orange"),name="HIV status") +
  scale_fill_manual(values=c("steelblue","orange"),name="HIV status") +
  guides(fill="none",lty="none",pch="none") +
  ggtitle("CIMT mean and 95% CI over time by participant group.") + 
  theme(text=element_text(size=16)) +
  xlab("Study visit (months since enrolment)") +
  ylab("CIMT (mm)") +
  xlim(c(0,24)) + 
  facet_wrap(~factor(physio.bmi_cat_enrolment,levels=c("underweight","normal","overweight","obese")),nrow = 2)

print(gCIMTbmi)

Figure 8.4: CIMT values, summarised by sample means and 95% confidence intervals, over time, by participant type and BMI group.

gCIMThyper<-tmpDatHyper %>%
  filter(visitNum==0 | visitNum==24) %>%
  ggplot(mapping=aes(x=visitNum,y=cimtMean,ymin=cimtMean-1.96*cimtSE,ymax=cimtMean+1.96*cimtSE,col=crf.PLWH,pch=crf.PLWH,lty=crf.PLWH,fill=crf.PLWH)) +
  geom_point(size=3) +
  geom_line(lwd=1) +
  geom_ribbon(alpha=0.25,linetype=0) +
  scale_color_manual(values=c("steelblue","orange"),name="HIV status") +
  scale_fill_manual(values=c("steelblue","orange"),name="HIV status") +
  guides(fill="none",lty="none",pch="none") +
  ggtitle("CIMT mean and 95% CI over time by participant group.") + 
  theme(text=element_text(size=16)) +
  xlab("Study visit (months since enrolment)") +
  ylab("CIMT (mm)") +
  xlim(c(0,24)) + 
  facet_wrap(~physio.hypertension_cat_enrolment,nrow = 1)

print(gCIMThyper)

Figure 8.5: CIMT values, summarised by sample means and 95% confidence intervals, over time, by participant type and hypertension group.

8.1.2 Regression models

CIMT is only measured at baseline and 24 months. For this reason, simple GLM models are fitted for CIMT. We use standard linear regression models for a change-from-baseline analysis (change-from-baseline CIMT is the response variable).

As for PWV, we fit 3 models: unadjusted model for all participants, adjusted model for all participants and an adjusted model for PLWH only. Baseline CIMT measurements are included as a predictor in all models as is recommended practice.

In the first model, the unadjusted model, we observe a statistically significant difference between PLWH and controls.

tmpLong<-mice::complete(rhiccaDatWide_impMice, action='long', include=TRUE) %>%
  dplyr::mutate(
      cmv.cmv_CFB=cmv.cmv_6months-cmv.cmv_enrolment,
      vzv.vzv_CFB=vzv.vzv_6months-vzv.vzv_enrolment
    )
rhiccaDatWide_impMice <- as.mids(tmpLong)


tmpDatWide<-rhiccaDatWide %>%
  mutate(
    pwvChange=pwv.pwv_24months-pwv.pwv_enrolment,
    cimtChange=cimt.cimt_24months-cimt.cimt_enrolment,
    abpiChange=abpi.abpi_24months-abpi.abpi_enrolment,
    egfrChange=eGFR_24months-eGFR_enrolment
  )

obj1GlmCimt.ptype<-glm(cimtChange~ifelse(crf.ptype=="Patient",1,0)+cimt.cimt_enrolment,data=tmpDatWide) # another marker of vasculopathy; unadjusted model

obj1GlmCimt.baseline.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~cimt.cimt_enrolment))
obj1GlmCimt.ptype.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~crf.ptype+cimt.cimt_enrolment))
obj1GlmCimt.sex.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~crf.gender+cimt.cimt_enrolment))
obj1GlmCimt.age.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~crf.age+cimt.cimt_enrolment))
obj1GlmCimt.waisthip.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~physio.waisthip_enrolment+cimt.cimt_enrolment))
obj1GlmCimt.bmi.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~physio.bmi_enrolment+cimt.cimt_enrolment))
obj1GlmCimt.diabp.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~physio.bp_dia_enrolment+cimt.cimt_enrolment))
obj1GlmCimt.sysbp.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~physio.bp_sys_enrolment+cimt.cimt_enrolment))
obj1GlmCimt.glucose.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~lims.Glucose_enrolment+cimt.cimt_enrolment))
obj1GlmCimt.cholesterol.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~lims.Cholesterol_enrolment+cimt.cimt_enrolment))
obj1GlmCimt.cmv.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~cmvStatus+cimt.cimt_enrolment))
obj1GlmCimt.vzv.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~vzvStatus+cimt.cimt_enrolment))

obj1GlmCimtAdjGlucChol<-glm(cimtChange~ifelse(crf.ptype=="Patient",1,0)+cimt.cimt_enrolment+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment,data=tmpDatWide) # another marker of vasculopathy; unadjusted model
obj1GlmCimtAdjGlucCholInt<-glm(cimtChange~crf.ptype+cimt.cimt_enrolment+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment*crf.gender+lims.Cholesterol_enrolment*crf.gender,data=tmpDatWide) # another marker of vasculopathy; unadjusted model

obj1GlmCimtAdjGlucChol.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~crf.ptype+cimt.cimt_enrolment+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus+vzvStatus)) 
#summary(pool(obj1GlmCimtAdjGlucChol.mid))

obj1GlmCimtAdjGlucChol.quantCmvVzv.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~ifelse(crf.ptype=="Patient",1,0)+cimt.cimt_enrolment+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmv.cmv_enrolment+cmv.cmv_CFB+vzv.vzv_enrolment+vzv.vzv_CFB)) # no sig assoc with VZV

obj1GlmCimtAdjCmvVzvOnly.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~ifelse(crf.ptype=="Patient",1,0)+cimt.cimt_enrolment+cmvStatus+vzvStatus))

obj1GlmCimtAdjGlucCholSolMarkers.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~ifelse(crf.ptype=="Patient",1,0)+cimt.cimt_enrolment+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+lum.ICAM1_enrolment+lum.CD163_enrolment+lum.IL6_enrolment)) 

obj1GlmCimtAdjGlucChol.noPtype.mid<-with(rhiccaDatWide_impMice,glm(cimtChange~cimt.cimt_enrolment+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment)) 

m<-length(obj1GlmCimtAdjGlucChol.mid$analyses)
obj1GlmCimtAdjGlucChol.mid.residuals.list<-list()
obj1GlmCimtAdjCmvVzvOnly.mid.residuals.list<-list()
obj1GlmCimtAdjGlucChol.noPtype.mid.residuals.list<-list()
for(i in 1:m){
  obj1GlmCimtAdjGlucChol.mid.residuals.list[[i]]<-residuals(obj1GlmCimtAdjGlucChol.mid$analyses[[i]])
  obj1GlmCimtAdjCmvVzvOnly.mid.residuals.list[[i]]<-residuals(obj1GlmCimtAdjCmvVzvOnly.mid$analyses[[i]])
  obj1GlmCimtAdjGlucChol.noPtype.mid.residuals.list[[i]]<-residuals(obj1GlmCimtAdjGlucChol.noPtype.mid$analyses[[i]])
  if(i==1){
    obj1GlmCimtAdjGlucChol.mid.residuals<-obj1GlmCimtAdjGlucChol.mid.residuals.list[[i]]/m
    obj1GlmCimtAdjCmvVzvOnly.mid.residuals<-obj1GlmCimtAdjCmvVzvOnly.mid.residuals.list[[i]]/m
    obj1GlmCimtAdjGlucChol.noPtype.mid.residuals<-obj1GlmCimtAdjGlucChol.noPtype.mid.residuals.list[[i]]/m
  }else{
    obj1GlmCimtAdjGlucChol.mid.residuals<-obj1GlmCimtAdjGlucChol.mid.residuals+obj1GlmCimtAdjGlucChol.mid.residuals.list[[i]]/m
    obj1GlmCimtAdjCmvVzvOnly.mid.residuals<-obj1GlmCimtAdjCmvVzvOnly.mid.residuals+obj1GlmCimtAdjCmvVzvOnly.mid.residuals.list[[i]]/m
    obj1GlmCimtAdjGlucChol.noPtype.mid.residuals<-obj1GlmCimtAdjGlucChol.noPtype.mid.residuals+obj1GlmCimtAdjGlucChol.noPtype.mid.residuals.list[[i]]/m
  }
}

tmpDatCimt<-data.frame(
  pid=rhiccaDatWide$pid,
  ptype=rhiccaDatWide$crf.ptype,
  cimtChangeResNoPtype=obj1GlmCimtAdjGlucChol.noPtype.mid.residuals,
  cimtChangeResPtype=obj1GlmCimtAdjGlucChol.mid.residuals,
  cimtChangeResPtypeCmvVzvOnly=obj1GlmCimtAdjCmvVzvOnly.mid.residuals
)

gCimtModAdjResiduals.noPtype<-tmpDatCimt %>%
  ggplot(mapping=aes(x=ptype,y=cimtChangeResNoPtype)) +
  geom_boxplot()

gCimtModAdjResiduals.Ptype<-tmpDatCimt %>%
  ggplot(mapping=aes(x=ptype,y=cimtChangeResPtype)) +
  geom_boxplot()

gCimtModAdjResiduals.PtypeCmvVzvOnly<-tmpDatCimt %>%
  ggplot(mapping=aes(x=ptype,y=cimtChangeResPtypeCmvVzvOnly)) +
  geom_boxplot()

#grid.arrange(gCimtModAdjResiduals.noPtype,gCimtModAdjResiduals.noPtype,nrow=1)

obj1GlmCimtAdjGlucCholSolMarkers<-glm(cimtChange~ifelse(crf.ptype=="Patient",1,0)+cimt.cimt_enrolment+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment*crf.gender+lims.Cholesterol_enrolment*crf.gender+lum.CD163_enrolment+lum.ICAM1_enrolment,data=tmpDatWide) # another marker of vasculopathy; unadjusted model

tmp<-mice::complete(rhiccaDatWide_impMice,1)
subsetHIV<-tmp$crf.ptype=="Patient"
rhiccaDatWide_impMice.hiv<-miceadds::subset_datlist(rhiccaDatWide_impMice,subset=subsetHIV,select=NULL,expr_subset=NULL,index=NULL)
rhiccaDatWide_impMice.hiv<-miceadds::datlist2mids(rhiccaDatWide_impMice.hiv)
tmpLong<-mice::complete(rhiccaDatWide_impMice.hiv, action='long', include=TRUE) %>%
  dplyr::mutate(
      cmv.cmv_CFB=cmv.cmv_6months-cmv.cmv_enrolment,
      vzv.vzv_CFB=vzv.vzv_6months-vzv.vzv_enrolment
    )
rhiccaDatWide_impMice.hiv <- as.mids(tmpLong)

obj1GlmCimtAdjGlucChol.hiv.mid<-with(rhiccaDatWide_impMice.hiv,glm(cimtChange~cimt.cimt_enrolment+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus+vzvStatus+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months))
obj1GlmCimtAdjGlucChol.quantCmvVzv.hiv.mid<-with(rhiccaDatWide_impMice.hiv,glm(cimtChange~cimt.cimt_enrolment+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmv.cmv_enrolment+cmv.cmv_CFB+vzv.vzv_enrolment+vzv.vzv_CFB+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months)) # no sig assoc with VZV
obj1GlmCimtAdjCmvVzvOnly.hiv.mid<-with(rhiccaDatWide_impMice.hiv,glm(cimtChange~cimt.cimt_enrolment+cmvStatus+vzvStatus))

summary(pool(obj1GlmCimt.mid)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(obj1GlmCimt.mid)))) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum,term)) %>%
  kableRn(caption="Model CIMT-1: Results for the linear regression model for change from baseline CIMT with participant type and baseline CIMT as predictors. Model fitted to multiply imputed data.",digits=4,row_names=c("Intercept","Participant type: PLWH","Baseline CIMT"),escape=F) %>%
  kable_styling(full_width=F)

In the adjusted model for all participants (adjusted for the same variables as the PWV models, but we use baseline values for all independent variables), this difference is no longer statistically significant. No other covariates have significant effects, though age is borderline above the significance threshold.

summary(pool(obj1GlmCimtAdjGlucChol.mid)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(obj1GlmCimtAdjGlucChol.mid)))) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum,term)) %>%
  kableRn(caption="Model CIMT-2: Results for the linear regression model for change from baseline CIMT with participant type, baseline CIMT and individual-level covariates (at baseline) as predictors. Model fitted to multiply imputed data.",digits=4,row_names=c("Intercept","Participant type: PLWH","Baseline CIMT","Sex: male","Age (in years)","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose","Cholesterol","CMV: progression to latency","CMV: reactivation","CMV: persistently high titres","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),escape=F) %>%
  kable_styling(full_width=F)

Likewise, in the PLWH only model, we do not observe any significant effects (age again borderline above the 5% significance threshold). This model also takes into account pre-existing and new onset vasculopathy (IRIS for the latter; we only consider 6 months after ART start). Vasculopathy is defined with respect to PWV (vasculopathy being defined as PWV>12).

summary(pool(obj1GlmCimtAdjGlucChol.hiv.mid)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(obj1GlmCimtAdjGlucChol.hiv.mid)))) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum,term)) %>%
  kableRn(caption="Model CIMT-3: Results for the linear mixed model for change from baseline CIMT with participant type, baseline CIMT and individual-level covariates (at baseline) as predictors. Model fitted to multiply imputed data.",digits=4,row_names=c("Intercept","Baseline CIMT","Sex: male","Age (in years)","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose","Cholesterol","CMV: progression to latency","CMV: reactivation","CMV: persistently high titres","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres","CD4 (baseline)","CD4 (change from Baseline at 24 months)","log10(viral load)"),escape=F) %>%
  kable_styling(full_width=F)

Table 8.1 shows a combined summary for the 3 CIMT models

#helper functions
sumFitandSum_cimt<-function(varString){
  mod<-with(rhiccaDatWide_impMice,glm(as.formula(paste(sep="","cimtChange~cimt.cimt_enrolment",varString))))
  
  sumObj<-summary(pool(mod)) %>%
    as.data.frame() %>%
    mutate(rownum=1:nrow(summary(pool(mod)))) %>%
    mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=3,round(digits=3,p.value)),bold=T),cell_spec(format(nsmall=3,round(digits=3,p.value)),bold=F))) %>%
    mutate(p.value=gsub(p.value,pattern="0.000",replacement="&lt;0.001"))
  
  return(sumObj)
}

sumFitandSum_cimt.hiv<-function(varString){
  mod<-with(rhiccaDatWide_impMice.hiv,glm(as.formula(paste(sep="","cimtChange~cimt.cimt_enrolment",varString))))
  
  sumObj<-summary(pool(mod)) %>%
    as.data.frame() %>%
    mutate(rownum=1:nrow(summary(pool(mod)))) %>%
    mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=3,round(digits=3,p.value)),bold=T),cell_spec(format(nsmall=3,round(digits=3,p.value)),bold=F))) %>%
    mutate(p.value=gsub(p.value,pattern="0.000",replacement="&lt;0.001"))
  
  return(sumObj)
}

tab1.baseline<-sumFitandSum_cimt("")
tab1.ptype<-sumFitandSum_cimt("+crf.ptype")
tab1.sex<-sumFitandSum_cimt("+crf.gender")
tab1.smoke<-sumFitandSum_cimt("+crf.smoke")
tab1.alcohol<-sumFitandSum_cimt("+crf.alcohol")
tab1.age<-sumFitandSum_cimt("+crf.age")
tab1.waisthip<-sumFitandSum_cimt("+physio.waisthip_enrolment")
tab1.bmi<-sumFitandSum_cimt("+physio.bmi_enrolment")
tab1.diabp<-sumFitandSum_cimt("+physio.bp_dia_enrolment")
tab1.sysbp<-sumFitandSum_cimt("+physio.bp_sys_enrolment")
tab1.glucose<-sumFitandSum_cimt("+lims.Glucose_enrolment")
tab1.cholesterol<-sumFitandSum_cimt("+lims.Cholesterol_enrolment")
tab1.ckd<-sumFitandSum_cimt("+eGFR_existingCKD")
tab1.cmv<-sumFitandSum_cimt("+cmvStatus")
tab1.vzv<-sumFitandSum_cimt("+vzvStatus")

tab1<-rbind(tab1.baseline,
            tab1.ptype,
            tab1.sex,
            tab1.age,
            tab1.smoke,
            tab1.alcohol,
            tab1.waisthip,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            tab1.ckd,
            tab1.cmv,
            tab1.vzv)

tab2<-sumFitandSum_cimt("+crf.ptype+crf.gender+crf.age+crf.smoke+crf.alcohol+physio.waisthip_enrolment+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+eGFR_existingCKD+cmvStatus+vzvStatus")

tab3<-sumFitandSum_cimt.hiv("+crf.gender+crf.age+crf.smoke+crf.alcohol+physio.waisthip_enrolment+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+eGFR_existingCKD+cmvStatus+vzvStatus+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")

tab4.baseline<-sumFitandSum_cimt.hiv("")
tab4.sex<-sumFitandSum_cimt.hiv("+crf.gender")
tab4.age<-sumFitandSum_cimt.hiv("+crf.age")
tab4.smoke<-sumFitandSum_cimt.hiv("+crf.smoke")
tab4.alcohol<-sumFitandSum_cimt.hiv("+crf.alcohol")
tab4.waisthip<-sumFitandSum_cimt.hiv("+physio.waisthip_enrolment")
tab4.bmi<-sumFitandSum_cimt.hiv("+physio.bmi_enrolment")
tab4.diabp<-sumFitandSum_cimt.hiv("+physio.bp_dia_enrolment")
tab4.sysbp<-sumFitandSum_cimt.hiv("+physio.bp_sys_enrolment")
tab4.glucose<-sumFitandSum_cimt.hiv("+lims.Glucose_enrolment")
tab4.cholesterol<-sumFitandSum_cimt.hiv("+lims.Cholesterol_enrolment")
tab4.ckd<-sumFitandSum_cimt.hiv("+eGFR_existingCKD")
tab4.cmv<-sumFitandSum_cimt.hiv("+cmvStatus")
tab4.vzv<-sumFitandSum_cimt.hiv("+vzvStatus")
tab4.cd4base<-sumFitandSum_cimt.hiv("+lims.CD4_Absolute_enrolment")
tab4.cd4change<-sumFitandSum_cimt.hiv("+lims.CD4_Absolute_changeFromBaselineAt24Months")
tab4.vl<-sumFitandSum_cimt.hiv("+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")

tab4<-rbind(tab4.baseline,
            tab4.sex,
            tab4.age,
            tab4.smoke,
            tab4.alcohol,
            tab4.waisthip,
            tab4.bmi,
            tab4.diabp,
            tab4.sysbp,
            tab4.glucose,
            tab4.cholesterol,
            tab4.ckd,
            tab4.cmv,
            tab4.vzv,
            tab4.cd4base,
            tab4.cd4change,
            tab4.vl)

allRows<-setdiff(unique(c(tab1$term,tab2$term,tab3$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,coefficient.unadj=NA,SE.unadj=NA,pvalue.unadj=NA,coefficient.adj=NA,SE.adj=NA,pvalue.adj=NA,coefficient.hivadj=NA,SE.hivadj=NA,pvalue.hivadj=NA)

tabAll$coefficient.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$SE.unadj<-tab1$std.error[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$coefficient.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$SE.adj<-tab2$std.error[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tabAll$coefficient.hivadj<-tab3$estimate[match(tabAll$term,tab3$term)]
tabAll$SE.hivadj<-tab3$std.error[match(tabAll$term,tab3$term)]
tabAll$pvalue.hivadj<-tab3$p.value[match(tabAll$term,tab3$term)]
                              
tab5<-summary(pool(obj1GlmCimtAdjCmvVzvOnly.mid)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(obj1GlmCimtAdjCmvVzvOnly.mid)))) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum,term))

tab6<-summary(pool(obj1GlmCimtAdjCmvVzvOnly.hiv.mid)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(obj1GlmCimtAdjCmvVzvOnly.hiv.mid)))) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum,term))

tab7<-summary(pool(obj1GlmCimtAdjGlucChol.quantCmvVzv.mid)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(obj1GlmCimtAdjGlucChol.quantCmvVzv.mid)))) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum,term))

tab8<-summary(pool(obj1GlmCimtAdjGlucChol.quantCmvVzv.hiv.mid)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(obj1GlmCimtAdjGlucChol.quantCmvVzv.hiv.mid)))) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum,term))

tab<-rbind(tab1,tab2,tab3)
tabAlt<-rbind(tab5,tab6)
tabAlt2<-rbind(tab7,tab8)

# rn<-c(
#     c("Intercept","Participant type: PLWH","Baseline CIMT"),
#     c("Intercept ","Participant type: PLWH ","Baseline CIMT ","Sex: male","Age (in years)","Waist-hip ratio","Diastolic blood pressure","Systolic blood pressure","Glucose","Cholesterol","CMV: progression to latency","CMV: reactivation","CMV: persistently high titres","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),
#     c("Intercept  ","Baseline CIMT  ","Sex: male ","Age (in years) ","Waist-hip ratio ","Diastolic blood pressure ","Systolic blood pressure ","Glucose ","Cholesterol ","CMV: progression to latency ","CMV: reactivation ","CMV: persistently high titres ","VZV: progression to latency ","VZV: reactivation ","VZV: persistently high titres ","CD4 (baseline)","CD4 (change from Baseline at 24 months)","log10(viral load)")
#   )

rn<-c("Baseline CIMT","Participant type: PLWH","Sex: male","Age (in years)","Current smoker","Drink alcohol","Waist-hip ratio","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose","Cholesterol","Existing CKD","CMV: progression to low titres","CMV: progression to high titres","CMV: persistently high titres","VZV: progression to low titres","VZV: progression to high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (change from Baseline at 24 months)","log10 viral load (baseline)")

rnAlt<-c(
    c("Intercept ","Participant type: PLWH ","Baseline CIMT ","CMV: progression to low titres ","CMV: progression to high titres ","CMV: persistently high titres ","VZV: progression to low titres ","VZV: progression to high titres ","VZV: persistently high titres "),
    c("Intercept  ","Baseline CIMT  ","CMV: progression to low titres  ","CMV: progression to high titres  ","CMV: persistently high titres  ","VZV: progression to low titres  ","VZV: progression to high titres  ","VZV: persistently high titres  ")
  )

rnAlt2<-c(
    c("Intercept ","Participant type: PLWH ","Baseline CIMT ","Sex: male","Age (in years)","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose","Cholesterol","CMV (enrolment)","CMV (change from baseline at 6 months)","VZV (enrolment)","VZV (change from baseline at 6 months)"),
    c("Intercept  ","Baseline CIMT  ","Sex: male ","Age (in years) ","BMI ","Diastolic blood pressure ","Systolic blood pressure ","Glucose ","Cholesterol ","CMV (enrolment) ","CMV (change from baseline at 6 months) ","VZV (enrolment) ","VZV (change from baseline at 6 months) ","CD4 (baseline)","CD4 (change from Baseline at 24 months)")
  )

options(knitr.kable.NA = '')

tabAll %>%
  dplyr::select(!term) %>%
  kableRn(digits=4,escape=F,row_names=rn,col.names=rep(c("Coefficient","SE","p"),3),caption="Model summaries for all models fitted to CIMT as outcome (unadjusted, single variable models for all participants, multiple regression model for all participants and multiple regression model for PLWH only.)") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  add_header_above(header=c(" "=1,"Unadjusted models, all data" = 3, "Multiple regression model, all data" = 3, "Multiple regression model, PLWH" = 3)) %>%
  kable_styling(full_width=F)

Table 8.1: Model summaries for all models fitted to CIMT as outcome (unadjusted, single variable models for all participants, multiple regression model for all participants and multiple regression model for PLWH only.)
	Unadjusted models, all data			Multiple regression model, all data			Multiple regression model, PLWH
	Coefficient	SE	p	Coefficient	SE	p	Coefficient	SE	p
Baseline CIMT	0.0427	0.0431	0.326	0.0085	0.0558	0.880	0.0558	0.0704	0.432
Participant type: PLWH	0.0164	0.0068	0.017	0.0098	0.0082	0.233
Sex: male	0.0016	0.0059	0.790	-0.0032	0.0068	0.636	-0.0044	0.0086	0.611
Age (in years)	0.0008	0.0006	0.178	0.0008	0.0006	0.179	0.0010	0.0008	0.195
Current smoker	-0.0043	0.0087	0.620	-0.0068	0.0105	0.515	-0.0071	0.0122	0.559
Drink alcohol	0.0005	0.0073	0.942	0.0025	0.0088	0.774	0.0026	0.0099	0.795
Waist-hip ratio	-0.0387	0.0435	0.374	-0.0434	0.0461	0.347	-0.0416	0.0547	0.447
BMI	-0.0010	0.0006	0.092	-0.0001	0.0007	0.886	-0.0006	0.0010	0.548
Diastolic blood pressure	-0.0002	0.0002	0.355	-0.0004	0.0005	0.476	-0.0005	0.0006	0.369
Systolic blood pressure	-0.0001	0.0001	0.555	0.0002	0.0003	0.494	0.0004	0.0004	0.290
Glucose	0.0002	0.0014	0.898	0.0005	0.0015	0.739	0.0011	0.0018	0.546
Cholesterol	-0.0046	0.0027	0.086	-0.0022	0.0030	0.467	-0.0044	0.0037	0.244
Existing CKD	-0.0087	0.0136	0.521	-0.0133	0.0138	0.339	-0.0184	0.0168	0.275
CMV: progression to low titres	-0.0081	0.0202	0.687	-0.0069	0.0205	0.736	-0.0011	0.0234	0.961
CMV: progression to high titres	0.0179	0.0082	0.030	0.0123	0.0084	0.143	0.0144	0.0098	0.142
CMV: persistently high titres	0.0114	0.0087	0.193	0.0071	0.0088	0.423	0.0074	0.0101	0.461
VZV: progression to low titres	0.0201	0.0097	0.041	0.0182	0.0097	0.062	0.0219	0.0120	0.071
VZV: progression to high titres	0.0049	0.0139	0.725	0.0052	0.0142	0.714	0.0064	0.0158	0.684
VZV: persistently high titres	0.0163	0.0078	0.037	0.0137	0.0079	0.083	0.0187	0.0093	0.046
CD4 (baseline)							0.0000	0.0000	0.592
CD4 (change from Baseline at 24 months)							0.0000	0.0000	0.241
log10 viral load (baseline)							0.0001	0.0029	0.966

write.csv(tabAll,row.names=F,file=paste(sep="_",outPrefix,"CIMT_regressionModels.csv"))

tabAll.cimt<-tabAll

Note that the intercept in the unadjusted model gives the average change in CIMT over 24 months:

sumFitandSum_cimt.hiv("-cimt.cimt_enrolment") %>%
  dplyr::select(!c(term,rownum,statistic,df)) %>%
  kableRn(digits=3,escape=F,row_names=c("(Intercept)"),col.names=c("Coefficient","SE","p"),caption="Model summary for the basic unadjusted model fitted to CIMT as outcome. Model fitted to PLWH only.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  kable_styling(full_width=F)

Table 8.2: Model summary for the basic unadjusted model fitted to CIMT as outcome. Model fitted to PLWH only.
	Coefficient	SE	p
(Intercept)	0.033	0.003	<0.001

We can also pull this out for the adjusted model:

tab3 %>%
  dplyr::filter(term=="(Intercept)") %>%
  dplyr::select(!c(term,rownum,statistic,df)) %>%
  kableRn(digits=3,escape=F,row_names=c("(Intercept)"),col.names=c("Coefficient","SE","p"),caption="Model summary (showing intercept only!) for the fully adjusted model fitted to CIMT as outcome. Model fitted to PLWH only.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  kable_styling(full_width=F)

Table 8.3: Model summary (showing intercept only!) for the fully adjusted model fitted to CIMT as outcome. Model fitted to PLWH only.
	Coefficient	SE	p
(Intercept)	-0.002	0.057	0.971

And we can repeat this for all participants. The intercept in the unadjusted model gives the average change in CIMT over 24 months:

sumFitandSum_cimt("-cimt.cimt_enrolment") %>%
  dplyr::select(!c(term,rownum,statistic,df)) %>%
  kableRn(digits=3,escape=F,row_names=c("(Intercept)"),col.names=c("Coefficient","SE","p"),caption="Model summary for the basic unadjusted model fitted to CIMT as outcome. Model fitted to all participants.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  kable_styling(full_width=F)

Table 8.4: Model summary for the basic unadjusted model fitted to CIMT as outcome. Model fitted to all participants.
	Coefficient	SE	p
(Intercept)	0.03	0.003	<0.001

We can also pull this out for the adjusted model:

tab2 %>%
  dplyr::filter(term=="(Intercept)") %>%
  dplyr::select(!c(term,rownum,statistic,df)) %>%
  kableRn(digits=3,escape=F,row_names=c("(Intercept)"),col.names=c("Coefficient","SE","p"),caption="Model summary (showing intercept only!) for the fully adjusted model fitted to CIMT as outcome. Model fitted to all participants.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  kable_styling(full_width=F)

Table 8.5: Model summary (showing intercept only!) for the fully adjusted model fitted to CIMT as outcome. Model fitted to all participants.
	Coefficient	SE	p
(Intercept)	0.025	0.047	0.588

tab4$term<-as.character(tab4$term)
tab3$term<-as.character(tab3$term)
allRows<-setdiff(unique(c(tab4$term,tab3$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,coefficient.unadj=NA,SE.unadj=NA,pvalue.unadj=NA,coefficient.adj=NA,SE.adj=NA,pvalue.adj=NA)

tabAll$coefficient.unadj<-tab4$estimate[match(tabAll$term,tab4$term)]
tabAll$SE.unadj<-tab4$std.error[match(tabAll$term,tab4$term)]
tabAll$pvalue.unadj<-tab4$p.value[match(tabAll$term,tab4$term)]

tabAll$coefficient.adj<-tab3$estimate[match(tabAll$term,tab3$term)]
tabAll$SE.adj<-tab3$std.error[match(tabAll$term,tab3$term)]
tabAll$pvalue.adj<-tab3$p.value[match(tabAll$term,tab3$term)]

rn<-c("Baseline CIMT","Sex: male","Age (in years)","Current smoker","Alcohol drinker","Waist-hip ratio","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose","Cholesterol","Existing CKD","CMV: progression to low titres","CMV: progression to high titres","CMV: persistently high titres","VZV: progression to low titres","VZV: progression to high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (change from Baseline at 24 months)","log10 viral load (baseline)")

options(knitr.kable.NA = '')

tabAll %>%
  dplyr::select(!term) %>%
  kableRn(digits=3,escape=F,row_names=rn,col.names=rep(c("Coefficient","SE","p"),2),caption="Model summaries for all models fitted to CIMT as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  add_header_above(header=c(" "=1,"Unadjusted models" = 3, "Multiple regression model" = 3)) %>%
  kable_styling(full_width=F)

Table 8.6: Model summaries for all models fitted to CIMT as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only.
	Unadjusted models			Multiple regression model
	Coefficient	SE	p	Coefficient	SE	p
Baseline CIMT	0.103	0.054	0.058	0.056	0.070	0.432
Sex: male	0.001	0.007	0.876	-0.004	0.009	0.611
Age (in years)	0.001	0.001	0.133	0.001	0.001	0.195
Current smoker	-0.003	0.010	0.742	-0.007	0.012	0.559
Alcohol drinker	-0.001	0.009	0.946	0.003	0.010	0.795
Waist-hip ratio	-0.031	0.052	0.548	-0.042	0.055	0.447
BMI	-0.001	0.001	0.222	-0.001	0.001	0.548
Diastolic blood pressure	0.000	0.000	0.693	-0.001	0.001	0.369
Systolic blood pressure	0.000	0.000	0.875	0.000	0.000	0.290
Glucose	0.001	0.002	0.611	0.001	0.002	0.546
Cholesterol	-0.005	0.004	0.205	-0.004	0.004	0.244
Existing CKD	-0.014	0.017	0.401	-0.018	0.017	0.275
CMV: progression to low titres	-0.008	0.023	0.716	-0.001	0.023	0.961
CMV: progression to high titres	0.017	0.009	0.062	0.014	0.010	0.142
CMV: persistently high titres	0.008	0.010	0.403	0.007	0.010	0.461
VZV: progression to low titres	0.022	0.012	0.061	0.022	0.012	0.071
VZV: progression to high titres	0.004	0.016	0.801	0.006	0.016	0.684
VZV: persistently high titres	0.020	0.009	0.032	0.019	0.009	0.046
CD4 (baseline)	0.000	0.000	0.668	0.000	0.000	0.592
CD4 (change from Baseline at 24 months)	0.000	0.000	0.335	0.000	0.000	0.241
log10 viral load (baseline)	0.001	0.003	0.664	0.000	0.003	0.966

tabAll.cimt.hiv<-tabAll

tabAlt %>%
  kableRn(digits=4,escape=F,row_names=rnAlt,caption="Model summaries for 2 additional models fitted to CIMT as outcome (adjusted model for all participants and PLWH only where adjustment is only for CMV and VZV status.)") %>%
  pack_rows(index = c("Adjusted model, all data" = 9, "Adjusted model, PLWH only" = 8)) %>%
  kable_styling(full_width=F)

Table 8.7: Model summaries for 2 additional models fitted to CIMT as outcome (adjusted model for all participants and PLWH only where adjustment is only for CMV and VZV status.)
	estimate	std.error	statistic	df	p.value
Adjusted model, all data
Intercept	-0.0096	0.0234	-0.4120	69.5976	0.6816
Participant type: PLWH	0.0113	0.0072	1.5655	407.2418	0.1183
Baseline CIMT	0.0456	0.0437	1.0432	57.9606	0.3012
CMV: progression to low titres	-0.0090	0.0200	-0.4523	550.4132	0.6512
CMV: progression to high titres	0.0140	0.0084	1.6718	535.8978	0.0952
CMV: persistently high titres	0.0070	0.0089	0.7876	63.5009	0.4338
VZV: progression to low titres	0.0178	0.0096	1.8513	151.1601	0.0661
VZV: progression to high titres	0.0037	0.0141	0.2629	262.2942	0.7928
VZV: persistently high titres	0.0140	0.0078	1.7923	324.8541	0.0740
Adjusted model, PLWH only
Intercept	-0.0248	0.0267	-0.9276	93.9517	0.3560
Baseline CIMT	0.0945	0.0539	1.7524	71.6226	0.0840
CMV: progression to low titres	-0.0062	0.0225	-0.2746	578.9454	0.7837
CMV: progression to high titres	0.0164	0.0093	1.7724	569.6967	0.0769
CMV: persistently high titres	0.0065	0.0096	0.6762	78.6618	0.5009
VZV: progression to low titres	0.0215	0.0116	1.8470	104.4557	0.0676
VZV: progression to high titres	0.0051	0.0157	0.3233	280.5830	0.7467
VZV: persistently high titres	0.0189	0.0092	2.0634	300.4709	0.0399

tabAlt2 %>%
  kableRn(digits=4,escape=F,row_names=rnAlt2,caption="Model summaries for another 2 additional models fitted to CIMT as outcome (adjusted model for all participants and PLWH only where adjustment is for the continuous CMV and VZV measurements.)") %>%
  pack_rows(index = c("Adjusted model, all data" = 14, "Adjusted model, PLWH only" = 15)) %>%
  kable_styling(full_width=F)

Table 8.8: Model summaries for another 2 additional models fitted to CIMT as outcome (adjusted model for all participants and PLWH only where adjustment is for the continuous CMV and VZV measurements.)
	estimate	std.error	statistic	df	p.value
Adjusted model, all data
Intercept	-0.0064	0.0323	-0.1990	196.7261	0.8425
Participant type: PLWH	0.0105	0.0082	1.2856	267.3073	0.1997
Baseline CIMT	0.0082	0.0559	0.1474	57.9949	0.8833
Sex: male	-0.0039	0.0064	-0.6116	290.6100	0.5412
Age (in years)	0.0007	0.0006	1.2439	34.8876	0.2218
BMI	-0.0002	0.0007	-0.3189	499.0811	0.7500
Diastolic blood pressure	-0.0003	0.0005	-0.6557	378.2575	0.5124
Systolic blood pressure	0.0002	0.0003	0.6019	419.3773	0.5476
Glucose	0.0001	0.0014	0.0944	792.1813	0.9248
Cholesterol	-0.0022	0.0030	-0.7437	154.2561	0.4582
CMV (enrolment)	0.0000	0.0000	0.1661	160.6307	0.8683
CMV (change from baseline at 6 months)	0.0000	0.0000	1.3049	286.1586	0.1930
VZV (enrolment)	0.0000	0.0000	1.3838	263.9358	0.1676
VZV (change from baseline at 6 months)	0.0000	0.0000	-0.3461	349.6715	0.7295
Adjusted model, PLWH only
Intercept	-0.0358	0.0381	-0.9396	212.3443	0.3485
Baseline CIMT	0.0544	0.0704	0.7728	54.1398	0.4430
Sex: male	-0.0049	0.0080	-0.6044	245.0394	0.5461
Age (in years)	0.0010	0.0008	1.2713	29.9475	0.2134
BMI	-0.0007	0.0009	-0.7287	236.2229	0.4669
Diastolic blood pressure	-0.0005	0.0006	-0.8766	393.4250	0.3813
Systolic blood pressure	0.0004	0.0004	1.0493	508.8965	0.2945
Glucose	0.0007	0.0017	0.4064	516.9907	0.6846
Cholesterol	-0.0043	0.0038	-1.1356	137.8255	0.2581
CMV (enrolment)	0.0000	0.0000	0.3024	188.6485	0.7627
CMV (change from baseline at 6 months)	0.0000	0.0000	1.4689	366.1948	0.1427
VZV (enrolment)	0.0000	0.0000	1.4658	199.3832	0.1443
VZV (change from baseline at 6 months)	0.0000	0.0000	-0.2293	256.0582	0.8188
CD4 (baseline)	0.0000	0.0000	0.6676	82.6753	0.5062
CD4 (change from Baseline at 24 months)	0.0000	0.0000	1.1613	92.6416	0.2485

8.2 Pulse Wave Velocity (PWV)

As an executive summary, the RHICCA data do not show that there is an association between PWV over time and participant type. Whilst a weak association with participant type is observed in an unadjusted model (both in terms of overall PWV level and change in PWV over time), this disappears when the model takes other cardiovascular risk factors into account. On average, PWV decreases over the 24 month study period.

8.2.1 Descriptive summaries

Figures 8.6, 8.7, 8.8, 8.9 and 8.10 summarise the distribution of PWV between participant types overall and at the different timepoints.

There is a marked dip in PWV in controls at 18 months and this is consistent: it can be seen overall, across all BMI groups (though perhaps fitting more in with the overall trend over time for overweight and obese) and across all blood pressure groups.

We investigated whether this dip could be due to observer variability. However, it does not seem to be explained by a staff member who would consistently record lower PWV. While 2 of the 3 staff with lowest average PWV measurements across all visits (excluding at 18 months), IS and SN, measured a slightly larger proportion of participants at 18 months (total of 79% compared to 60%), the dip persists when the plot is redone with excluding measurements from these 2 staff (though the dip is less pronounced; figure not shown).

rhiccaDatLong %>%
  as.data.frame() %>%
  filter(!is.na(crf.PLWH)) %>%
  histFun(var="pwv.pwv",groupVar="crf.PLWH",yLab="PWV (m/s)",labels=c("Control","PLWH"),title="PWV (all visits)")

Figure 8.6: Histogram of all PWV measurements across all visits, overall and by participant type.

g1<-rhiccaDatWide %>%
  as.data.frame() %>%
  filter(!is.na(crf.ptype)) %>%
  violinPlotFun(var="pwv.pwv_enrolment",groupVar="crf.PLWH",yLab="PWV (m/s)",labels=c("Control","PLWH"),ylim=c(3,20),title="PWV at baseline")

g6<-rhiccaDatWide %>%
  as.data.frame() %>%
  filter(!is.na(crf.ptype)) %>%
  violinPlotFun(var="pwv.pwv_6months",groupVar="crf.PLWH",yLab="PWV (m/s)",labels=c("Control","PLWH"),ylim=c(3,20),title="PWV at 6 months")

g12<-rhiccaDatWide %>%
  as.data.frame() %>%
  filter(!is.na(crf.ptype)) %>%
  violinPlotFun(var="pwv.pwv_12months",groupVar="crf.PLWH",yLab="PWV (m/s)",labels=c("Control","PLWH"),ylim=c(3,20),title="PWV at 12 months")

g18<-rhiccaDatWide %>%
  as.data.frame() %>%
  filter(!is.na(crf.ptype)) %>%
  violinPlotFun(var="pwv.pwv_18months",groupVar="crf.PLWH",yLab="PWV (m/s)",labels=c("Control","PLWH"),ylim=c(3,20),title="PWV at 18 months")

g24<-rhiccaDatWide %>%
  as.data.frame() %>%
  filter(!is.na(crf.ptype)) %>%
  violinPlotFun(var="pwv.pwv_24months",groupVar="crf.PLWH",yLab="PWV (m/s)",labels=c("Control","PLWH"),ylim=c(3,20),title="PWV at 24 months")

g30<-rhiccaDatWide %>%
  as.data.frame() %>%
  filter(!is.na(crf.ptype)) %>%
  violinPlotFun(var="pwv.pwv_30months",groupVar="crf.PLWH",yLab="PWV (m/s)",labels=c("Control","PLWH"),ylim=c(3,20),title="PWV at 30 months")

grid.arrange(g1,g6,g12,g18,g24,g30,nrow=2)

Figure 8.7: Violin plots of PWV measurements at the different visits, stratified by participant type.

gPWV<-tmpDatAll %>%
  ggplot(mapping=aes(x=visitNum,y=pwvMean,ymin=pwvMean-1.96*pwvSE,ymax=pwvMean+1.96*pwvSE,col=crf.PLWH,pch=crf.PLWH,lty=crf.PLWH,fill=crf.PLWH)) +
  geom_point(size=3) +
  geom_line(lwd=1) +
  geom_ribbon(alpha=0.25,linetype=0) +
  scale_color_manual(values=c("steelblue","orange"),name="HIV status") +
  scale_fill_manual(values=c("steelblue","orange"),name="HIV status") +
  guides(fill="none",lty="none",pch="none") +
  ggtitle("PWV mean and 95% CI over time by participant group.") + 
  theme(text=element_text(size=16)) +
  xlab("Study visit (months since enrolment)") +
  ylab("PWV (m/s)") +
  xlim(c(0,24))

tmpDatAllAdj<-rhiccaDatLong %>%
  filter(visitNum==0 | visitNum==6 | visitNum==12 | visitNum==18 | visitNum==24) %>%
  dplyr::select(pid,visitNum,crf.PLWH,pwv.pwv,physio.bmi_cat_enrolment,physio.hypertension_cat_enrolment) %>%
  mutate(
    physio.bmi_enrolment=rhiccaDatWide$physio.bmi_enrolment[match(pid,rhiccaDatWide$pid)],
    physio.waisthip_enrolment=rhiccaDatWide$physio.waisthip_enrolment[match(pid,rhiccaDatWide$pid)],
    physio.bp_dia_enrolment=rhiccaDatWide$physio.bp_dia_enrolment[match(pid,rhiccaDatWide$pid)],
    physio.bp_sys_enrolment=rhiccaDatWide$physio.bp_sys_enrolment[match(pid,rhiccaDatWide$pid)],
    lims.Glucose_enrolment=rhiccaDatWide$lims.Glucose_enrolment[match(pid,rhiccaDatWide$pid)],
    lims.Cholesterol_enrolment=rhiccaDatWide$lims.Cholesterol_enrolment[match(pid,rhiccaDatWide$pid)]
  ) %>%
  filter(rowSums(is.na(.))==0)
tmpDatAllAdj<-tmpDatAllAdj[order(tmpDatAllAdj$pid),]

geeModPwv<-geeglm(pwv.pwv~physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment,data=tmpDatAllAdj,id=factor(tmpDatAllAdj$pid),corstr="exchangeable")

tmpDatAllAdj<-tmpDatAllAdj %>%
  mutate(
    pwv.geeResid=residuals(geeModPwv)
  ) %>%
  group_by(visitNum,crf.PLWH) %>%
  summarise(pwvMean=mean(pwv.pwv,na.rm=T),pwvSE=sd(pwv.pwv,na.rm=T)/sqrt(sum(!is.na(pwv.pwv))),
            pwvAdjMean=mean(pwv.geeResid,na.rm=T),pwvAdjSE=sd(pwv.geeResid,na.rm=T)/sqrt(sum(!is.na(pwv.geeResid))),
            .groups="drop") 

gPWVAdj<-tmpDatAllAdj %>%
  filter(visitNum==0 | visitNum==6 | visitNum==12 | visitNum==18 | visitNum==24) %>%
  ggplot(mapping=aes(x=visitNum,y=pwvAdjMean,ymin=pwvAdjMean-1.96*pwvAdjSE,ymax=pwvAdjMean+1.96*pwvAdjSE,col=crf.PLWH,pch=crf.PLWH,lty=crf.PLWH,fill=crf.PLWH)) +
  geom_point(size=3) +
  geom_line(lwd=1) +
  geom_ribbon(alpha=0.25,linetype=0) +
  scale_color_manual(values=c("steelblue","orange"),name="HIV status") +
  scale_fill_manual(values=c("steelblue","orange"),name="HIV status") +
  guides(fill="none",lty="none",pch="none") +
  ggtitle("Adjusted PWV mean and 95% CI over time by participant group.") + 
  theme(text=element_text(size=16)) +
  xlab("Study visit (months since enrolment)") +
  ylab("Residual PWV (m/s)") +
  labs(caption="Longitudinal PWV has been adjusted for baseline CVD risk factors\n(waist-hip ratio, systolic and diastolic blood pressure, glucose, cholesterol)\nusing a GEE model with exchangeable correlation structure.") +
  xlim(c(0,24))

gPWVAdj.notitle<-tmpDatAllAdj %>%
  filter(visitNum==0 | visitNum==6 | visitNum==12 | visitNum==18 | visitNum==24) %>%
  ggplot(mapping=aes(x=visitNum,y=pwvAdjMean,ymin=pwvAdjMean-1.96*pwvAdjSE,ymax=pwvAdjMean+1.96*pwvAdjSE,col=crf.PLWH,pch=crf.PLWH,lty=crf.PLWH,fill=crf.PLWH)) +
  geom_point(size=3) +
  geom_line(lwd=1) +
  geom_ribbon(alpha=0.25,linetype=0) +
  scale_color_manual(values=c("steelblue","orange"),name="HIV status") +
  scale_fill_manual(values=c("steelblue","orange"),name="HIV status") +
  guides(fill="none",lty="none",pch="none",col="none") +
  theme(text=element_text(size=12),legend.position="left") +
  xlab("Study visit (months since enrolment)") +
  ylab("Residual PWV (m/s)") +
  #labs(caption="Longitudinal PWV has been adjusted for baseline CVD risk factors\n(waist-hip ratio, systolic and diastolic blood pressure, glucose, cholesterol)\nusing a GEE model with exchangeable correlation structure.") +
  xlim(c(0,24))

#print(gPWV)
grid.arrange(gPWV,gPWVAdj,nrow=1)

Figure 8.8: PWV sample mean and 95% CI over time by participant type.

gPWVbmi<-tmpDatBmi %>%
  ggplot(mapping=aes(x=visitNum,y=pwvMean,ymin=pwvMean-1.96*pwvSE,ymax=pwvMean+1.96*pwvSE,col=crf.PLWH,pch=crf.PLWH,lty=crf.PLWH,fill=crf.PLWH)) +
  geom_point(size=3) +
  geom_line(lwd=1) +
  geom_ribbon(alpha=0.25,linetype=0) +
  scale_color_manual(values=c("steelblue","orange"),name="HIV status") +
  scale_fill_manual(values=c("steelblue","orange"),name="HIV status") +
  guides(fill="none",lty="none",pch="none") +
  ggtitle("PWV mean and 95% CI over time by participant group.\n(Stratified by BMI.)") + 
  theme(text=element_text(size=16)) +
  xlab("Study visit (months since enrolment)") +
  ylab("PWV (m/s)") +
  xlim(c(0,24)) +
  facet_wrap(~factor(physio.bmi_cat_enrolment,levels=c("underweight","normal","overweight","obese")),nrow=2)

print(gPWVbmi)

Figure 8.9: PWV sample mean and 95% CI over time by participant type and BMI group.

gPWVhyper<-tmpDatHyper %>%
  ggplot(mapping=aes(x=visitNum,y=pwvMean,ymin=pwvMean-1.96*pwvSE,ymax=pwvMean+1.96*pwvSE,col=crf.PLWH,pch=crf.PLWH,lty=crf.PLWH,fill=crf.PLWH)) +
  geom_point(size=3) +
  geom_line(lwd=1) +
  geom_ribbon(alpha=0.25,linetype=0) +
  scale_color_manual(values=c("steelblue","orange"),name="HIV status") +
  scale_fill_manual(values=c("steelblue","orange"),name="HIV status") +
  guides(fill="none",lty="none",pch="none") +
  ggtitle("PWV mean and 95% CI over time by participant group.\n(Stratified by baseline hypertension group.)") + 
  theme(text=element_text(size=16)) +
  xlab("Study visit (months since enrolment)") +
  ylab("PWV (m/s)") +
  xlim(c(0,24)) +
  facet_wrap(~physio.hypertension_cat_enrolment,nrow=1)

print(gPWVhyper)

Figure 8.10: PWV sample mean and 95% CI over time by participant type and hypertension group.

8.2.2 Regression models

There are 3 models below and all are longitudinal, linear mixed models. The input data is multiply imputed data using data from the enrolments visit and visits at 6, 12, 18 and 24 months. Variables that are not immutable participant characteristics but for which only the baseline measurement has been used are indicated as such.

The first model is an an unadjusted model with only time (since baseline visits, in months) and participant type as predictors. This aims to summarise the data, but can be confounded by modifiable risk factors that affect both PWV and that are differently distributed among participant types. In this model both time and participant type are statistically significant, meaning there is evidence to suggest that PWV increased with time and is, on average, lower in PLWH compared to controls.

tmpDatWide<-rhiccaDatWide %>%
  mutate(
    pwvChange=pwv.pwv_24months-pwv.pwv_enrolment,
    cimtChange=cimt.cimt_24months-cimt.cimt_enrolment,
    abpiChange=abpi.abpi_24months-abpi.abpi_enrolment,
    egfrChange=eGFR_24months-eGFR_enrolment
  )

# LMMs - adjusted and unadjusted
obj1LmmPwv<-lmer(pwv.pwv~visitNumExactPwv+crf.ptype+(1|pid),data=rhiccaDatLong)

# obj1LmmPwv.timeonly.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum+(1|pid)))
# obj1LmmPwv.ptype.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum+crf.ptype+(1|pid)))
# obj1LmmPwv.sex.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum+crf.gender+(1|pid)))
# obj1LmmPwv.age.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum+crf.age+(1|pid)))
# obj1LmmPwv.waisthip.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum+physio.waisthip_enrolment+(1|pid)))
# obj1LmmPwv.diabp.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum+physio.bp_dia+(1|pid)))
# obj1LmmPwv.sysbp.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum+physio.bp_sys+(1|pid)))
# obj1LmmPwv.glucose.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum+lims.Glucose_enrolment+(1|pid)))
# obj1LmmPwv.cholesterol.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum+lims.Cholesterol_enrolment+(1|pid)))
# obj1LmmPwv.cmv.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum+cmvStatus+(1|pid)))
# obj1LmmPwv.vzv.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum+vzvStatus+(1|pid)))

rhiccaDatLong$physio.bmi_enrolment<-rhiccaDatWide$physio.bmi_enrolment[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$physio.waisthip_enrolment<-rhiccaDatWide$physio.waisthip_enrolment[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$lims.Glucose_enrolment<-rhiccaDatWide$lims.Glucose_enrolment[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$lims.Cholesterol_enrolment<-rhiccaDatWide$lims.Cholesterol_enrolment[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$lum.CD163_enrolment<-rhiccaDatWide$lum.CD163_enrolment[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$lum.CD163_6months<-rhiccaDatWide$lum.CD163_6months[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$lum.CD163_change<-(rhiccaDatWide$lum.CD163_6months-rhiccaDatWide$lum.CD163_enrolment)[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$lum.IL6_enrolment<-rhiccaDatWide$lum.IL6_enrolment[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$lum.IL6_6months<-rhiccaDatWide$lum.IL6_6months[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$lum.IL6_change<-(rhiccaDatWide$lum.IL6_6months-rhiccaDatWide$lum.IL6_enrolment)[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$lum.ICAM1_enrolment<-rhiccaDatWide$lum.ICAM1_enrolment[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$lum.ICAM1_6months<-rhiccaDatWide$lum.ICAM1_6months[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$lum.ICAM1_change<-(rhiccaDatWide$lum.ICAM1_6months-rhiccaDatWide$lum.ICAM1_enrolment)[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]
rhiccaDatLong$crf.artRegimen.2levels<-rhiccaDatWide$crf.artRegimen.2levels[match(rhiccaDatLong$pid,rhiccaDatWide$pid)]

obj1LmmPwvAdjGlucChol<-lmerTest::lmer(pwv.pwv~visitNum*crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia+physio.bp_sys+lims.Glucose_enrolment+lims.Cholesterol_enrolment+(1|pid),data=rhiccaDatLong)
obj1LmmPwvAdjGlucChol.mid<-with(rhiccaDatLong_impMice,lme4::lmer(pwv.pwv~visitNum*crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia+physio.bp_sys+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus+vzvStatus+(1|pid)))

tmp<-mice::complete(rhiccaDatLong_impMice,1)
subsetHIV<-(tmp$crf.ptype=="Patient")
rhiccaDatLong_impMice.hiv<-miceadds::subset_datlist(rhiccaDatLong_impMice,subset=subsetHIV,select=NULL,expr_subset=NULL,index=NULL)
rhiccaDatLong_impMice.hiv<-miceadds::datlist2mids(rhiccaDatLong_impMice.hiv)
obj1LmmPwvAdjGlucChol.hiv.mid<-with(rhiccaDatLong_impMice.hiv,lme4::lmer(pwv.pwv~visitNum+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia+physio.bp_sys+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus+vzvStatus+lims.CD4_Absolute_enrolment+lims.CD4_Absolute+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)+(1|pid)))

summary(pool(obj1LmmPwv.mid)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(obj1LmmPwv.mid)))) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum,term)) %>%
  kableRn(caption="Model PWV-1: Results for the linear mixed model for PWV with time and participant type as fixed factors, subject ID as random factor (random intercept only). Model fitted to multiply imputed data. Except for immutable participant characteristics, each variable is input longitudinally into the model with data from visits at enrolment, 6, 12, 18 and 24 month unless otherwise indicated.",digits=4,row_names=c("Intercept","Time since baseline visit (in months)","Participant type: PLWH"),escape=F) %>%
  kable_styling(full_width=F)

The second model is adjusted for participant characteristics and modifiable risk factors (sex, age, baseline waist-hip ratio, diastolic and systolic blood pressure, baseline glucose and baseline cholesterol). Further this model allows for a different trend over time between PLWH and controls. We note that while the overall trend over time is now weakened, it is still statistically significant and there is no evidence for a different trend over time between PLWH and control. The overall difference in PWV observed between PLWH and controls is no longer statistically significant and appears to be explained by age, diastolic and systolic blood pressure and baseline glucose values.

summary(pool(obj1LmmPwvAdjGlucChol.mid)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(obj1LmmPwvAdjGlucChol.mid)))) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum,term)) %>%
  kableRn(caption="Model PWV-2: Results for the linear mixed model for PWV with time, participant type and individual-level covariates as fixed factors (including glucose and cholesterol), subject ID as random factor (random intercept only). Model fitted to multiply imputed data. Except for immutable participant characteristics, each variable is input longitudinally into the model with data from visits at enrolment, 6, 12, 18 and 24 month unless otherwise indicated.",digits=4,row_names=c("Intercept","Time since baseline visit (in months)","Participant type: PLWH","Sex male","Age (in years)","BMI (baseline)","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","CMV: progression to latency","CMV: reactivation","CMV: persistently high titres","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres","Change in time since baseline coefficient in PLWH"),escape=F) %>%
  kable_styling(full_width=F)

The third model is fit to data from PLWH only. It is the same model as model 2 except that additionally CD4 count and log viral load are included as predictors. As viral load is near completely suppressed by 6 months, we only use baseline viral loads. CD4 counts are measured at baseline, 6, 12 and 24 months (note that this model therefore does not use data from the 18 month visit). We include both the baseline measurement and the change from baseline in the model.

There is again a significant decrease in PWV over time and age, diastolic and systolic blood pressure and glucose have an impact on PWV as in the previous model. However there is also a strongly statistically significant decrease in PWV associatied with increasing CD4 count.

summary(pool(obj1LmmPwvAdjGlucChol.hiv.mid)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(obj1LmmPwvAdjGlucChol.hiv.mid)))) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum,term)) %>%
  kableRn(caption="Model PWV-3: Results for the linear mixed model for PWV in PLWH only with subject ID as random factor (random intercept only). Model fitted to multiply imputed data. Except for immutable participant characteristics, each variable is input longitudinally into the model with data from visits at enrolment, 6, 12 and 24 month unless otherwise indicated.",digits=4,row_names=c("Intercept","Time since baseline visit (in months)","Sex male","Age (in years)","BMI (baseline)","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","CMV: progression to latency","CMV: reactivation","CMV: persistently high titres","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres","CD4 absolute (baseline)","CD4 absolute (change from baseline)","log10(viral load) at baseline"),escape=F) %>%
  kable_styling(full_width=F)

Table 8.9 shows a combined summary for the 3 PWV models

#helper functions
sumFitandSum_pwv<-function(varString){
  mod<-with(rhiccaDatLong_impMice,lme4::lmer(as.formula(paste(sep="+","pwv.pwv~(1|pid)",varString))))
  
  sumObj<-summary(pool(mod)) %>%
    as.data.frame() %>%
    mutate(rownum=1:nrow(summary(pool(mod)))) %>%
    mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F)))
  
  return(sumObj)
}

sumFitandSum_pwv.hiv<-function(varString){
  mod<-with(rhiccaDatLong_impMice.hiv,lme4::lmer(as.formula(paste(sep="+","pwv.pwv~(1|pid)",varString))))
  
  sumObj<-summary(pool(mod)) %>%
    as.data.frame() %>%
    mutate(rownum=1:nrow(summary(pool(mod)))) %>%
    mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F)))
  
  return(sumObj)
}

# tab1<-summary(pool(obj1LmmPwv.mid)) %>%
#   as.data.frame() %>%
#   mutate(rownum=1:nrow(summary(pool(obj1LmmPwv.mid)))) %>%
#   mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
#   dplyr::select(!c(rownum,term))
tab1.timeonly<-sumFitandSum_pwv("visitNum")
tab1.ptype<-sumFitandSum_pwv("visitNum+crf.ptype")
tab1.sex<-sumFitandSum_pwv("visitNum+crf.gender")
tab1.age<-sumFitandSum_pwv("visitNum+crf.age")
tab1.smoke<-sumFitandSum_pwv("visitNum+crf.smoke")
tab1.alcohol<-sumFitandSum_pwv("visitNum+crf.alcohol")
tab1.waisthip<-sumFitandSum_pwv("visitNum+physio.waisthip_enrolment")
tab1.bmi<-sumFitandSum_pwv("visitNum+physio.bmi_enrolment")
tab1.diabp<-sumFitandSum_pwv("visitNum+physio.bp_dia")
tab1.sysbp<-sumFitandSum_pwv("visitNum+physio.bp_sys")
tab1.glucose<-sumFitandSum_pwv("visitNum+lims.Glucose_enrolment")
tab1.cholesterol<-sumFitandSum_pwv("visitNum+lims.Cholesterol_enrolment")
tab1.ckd<-sumFitandSum_pwv("visitNum+eGFR_existingCKD")
tab1.cmv<-sumFitandSum_pwv("visitNum+cmvStatus")
tab1.vzv<-sumFitandSum_pwv("visitNum+vzvStatus")
tab1<-rbind(tab1.timeonly,
            tab1.ptype,
            tab1.sex,
            tab1.age,
            tab1.smoke,
            tab1.alcohol,
            tab1.waisthip,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            tab1.ckd,
            tab1.cmv,
            tab1.vzv)

tab2<-sumFitandSum_pwv("visitNum*crf.ptype+crf.gender+crf.age+crf.smoke+crf.alcohol+physio.waisthip_enrolment+physio.bmi_enrolment+physio.bp_dia+physio.bp_sys+lims.Glucose_enrolment+lims.Cholesterol_enrolment+eGFR_existingCKD+cmvStatus+vzvStatus")

tab3<-sumFitandSum_pwv.hiv("visitNum+crf.gender+crf.age+crf.smoke+crf.alcohol+physio.waisthip_enrolment+physio.bmi_enrolment+physio.bp_dia+physio.bp_sys+lims.Glucose_enrolment+lims.Cholesterol_enrolment+eGFR_existingCKD+cmvStatus+vzvStatus+lims.CD4_Absolute_enrolment+lims.CD4_Absolute+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")

tab4.timeonly<-sumFitandSum_pwv.hiv("visitNum")
tab4.sex<-sumFitandSum_pwv.hiv("visitNum+crf.gender")
tab4.age<-sumFitandSum_pwv.hiv("visitNum+crf.age")
tab4.smoke<-sumFitandSum_pwv.hiv("visitNum+crf.smoke")
tab4.alcohol<-sumFitandSum_pwv.hiv("visitNum+crf.alcohol")
tab4.waisthip<-sumFitandSum_pwv.hiv("visitNum+physio.waisthip_enrolment")
tab4.bmi<-sumFitandSum_pwv.hiv("visitNum+physio.bmi_enrolment")
tab4.diabp<-sumFitandSum_pwv.hiv("visitNum+physio.bp_dia")
tab4.sysbp<-sumFitandSum_pwv.hiv("visitNum+physio.bp_sys")
tab4.glucose<-sumFitandSum_pwv.hiv("visitNum+lims.Glucose_enrolment")
tab4.cholesterol<-sumFitandSum_pwv.hiv("visitNum+lims.Cholesterol_enrolment")
tab4.ckd<-sumFitandSum_pwv.hiv("visitNum+eGFR_existingCKD")
tab4.cmv<-sumFitandSum_pwv.hiv("visitNum+cmvStatus")
tab4.vzv<-sumFitandSum_pwv.hiv("visitNum+vzvStatus")
tab4.cd4base<-sumFitandSum_pwv.hiv("visitNum+lims.CD4_Absolute_enrolment")
tab4.cd4change<-sumFitandSum_pwv.hiv("visitNum+lims.CD4_Absolute")
tab4.vl<-sumFitandSum_pwv.hiv("visitNum+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")
tab4<-rbind(tab4.timeonly,
            tab4.sex,
            tab4.age,
            tab4.smoke,
            tab4.alcohol,
            tab4.waisthip,
            tab4.bmi,
            tab4.diabp,
            tab4.sysbp,
            tab4.glucose,
            tab4.cholesterol,
            tab4.ckd,
            tab4.cmv,
            tab4.vzv,
            tab4.cd4base,
            tab4.cd4change,
            tab4.vl
)

allRows<-setdiff(unique(c(tab1$term,tab2$term,tab3$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,coefficient.unadj=NA,SE.unadj=NA,pvalue.unadj=NA,coefficient.adj=NA,SE.adj=NA,pvalue.adj=NA,coefficient.hivadj=NA,SE.hivadj=NA,pvalue.hivadj=NA)

tabAll$coefficient.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$SE.unadj<-tab1$std.error[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$coefficient.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$SE.adj<-tab2$std.error[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tabAll$coefficient.hivadj<-tab3$estimate[match(tabAll$term,tab3$term)]
tabAll$SE.hivadj<-tab3$std.error[match(tabAll$term,tab3$term)]
tabAll$pvalue.hivadj<-tab3$p.value[match(tabAll$term,tab3$term)]

tabAll<-tabAll

# rn<-c(
#     c("Intercept","Time since baseline visit (in months)","Participant type: PLWH"),
#     c("Intercept ","Time since baseline visit (in months) ","Participant type: PLWH ","Sex male","Age (in years)","Waist-hip ratio (baseline)","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","CMV: progression to latency","CMV: reactivation","CMV: persistently high titres","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres","Change in time since baseline coefficient in PLWH"),
#     c("Intercept  ","Time since baseline visit (in months)  ","Sex male ","Age (in years) ","Waist-hip ratio (baseline) ","Diastolic blood pressure ","Systolic blood pressure ","Glucose (baseline) ","Cholesterol (baseline) ","CMV: progression to latency ","CMV: reactivation ","CMV: persistently high titres ","VZV: progression to latency ","VZV: reactivation ","VZV: persistently high titres ","CD4 absolute (baseline)","CD4 absolute (change from baseline)","log10(viral load) (baseline)")
#   )
# 
# tab %>%
#   kableRn(digits=4,escape=F,row_names=rn,caption="Model summaries for all 3 models fitted to PWV as outcome (unadjusted model, adjusted model and adjusted model for PLWH only.)") %>%
#   pack_rows(index = c("Unadjusted model, all data" = 3, "Adjusted model, all data" = 17, "Adjusted model, PLWH only" = 18)) %>%
#   kable_styling(full_width=F)

rn<-c("Time since baseline visit (in months)","Participant type: PLWH","Sex: male","Age (in years)","Current smoker","Alcohol drinker","Waist-hip ratio","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose","Cholesterol","Existing CKD","CMV: progression to low titres","CMV: progression to high titres","CMV: persistently high titres","VZV: progression to low titres","VZV: progression to high titres","VZV: persistently high titres","Change in time since baseline coefficient in PLWH","CD4 (baseline)","CD4 (longitudinal)","log10 viral load (baseline)")

tabAll %>%
  dplyr::select(!term) %>%
  kableRn(digits=4,escape=F,row_names=rn,col.names=rep(c("Coefficient","SE","p"),3),caption="Model summaries for all mixed effects models fitted to PWV as outcome (unadjusted, single variable models for all participants, multiple regression model for all participants and multiple regression model for PLWH only.)") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  add_header_above(header=c(" "=1,"Unadjusted models, all data" = 3, "Multiple regression model, all data" = 3, "Multiple regression model, PLWH" = 3)) %>%
  kable_styling(full_width=F)

Table 8.9: Model summaries for all mixed effects models fitted to PWV as outcome (unadjusted, single variable models for all participants, multiple regression model for all participants and multiple regression model for PLWH only.)
	Unadjusted models, all data			Multiple regression model, all data			Multiple regression model, PLWH
	Coefficient	SE	p	Coefficient	SE	p	Coefficient	SE	p
Time since baseline visit (in months)	-0.0251	0.0027	0.0000	-0.0168	0.0059	0.0043	-0.0139	0.0040	0.0007
Participant type: PLWH	-0.2428	0.1212	0.0454	0.1003	0.1351	0.4578
Sex: male	0.1025	0.1015	0.3130	-0.0523	0.1052	0.6198	-0.0908	0.1192	0.4473
Age (in years)	0.0778	0.0061	0.0000	0.0630	0.0058	0.0000	0.0603	0.0071	0.0000
Current smoker	0.0728	0.1639	0.6573	0.0371	0.1561	0.8122	0.0418	0.1662	0.8015
Alcohol drinker	0.0478	0.1225	0.6967	0.1466	0.1241	0.2383	0.1280	0.1313	0.3300
Waist-hip ratio	2.0841	0.7452	0.0052	-0.7283	0.6481	0.2614	-0.7116	0.7293	0.3296
BMI	0.0101	0.0101	0.3180	-0.0100	0.0101	0.3211	-0.0096	0.0120	0.4233
Diastolic blood pressure	0.0472	0.0029	0.0000	0.0280	0.0050	0.0000	0.0324	0.0063	0.0000
Systolic blood pressure	0.0317	0.0019	0.0000	0.0136	0.0034	0.0001	0.0106	0.0041	0.0110
Glucose	0.0716	0.0257	0.0054	0.0493	0.0212	0.0199	0.0551	0.0232	0.0174
Cholesterol	0.1132	0.0446	0.0115	0.0094	0.0428	0.8259	-0.0011	0.0484	0.9813
Existing CKD	0.3579	0.2586	0.1677	0.0255	0.2131	0.9047	0.0254	0.2329	0.9132
CMV: progression to low titres	-0.2632	0.3680	0.4748	-0.0776	0.3017	0.7971	-0.2147	0.3105	0.4894
CMV: progression to high titres	0.0982	0.1594	0.5384	0.1211	0.1398	0.3883	0.0841	0.1431	0.5572
CMV: persistently high titres	0.0093	0.1325	0.9439	0.0776	0.1150	0.5004	0.0306	0.1276	0.8106
VZV: progression to low titres	0.0925	0.1802	0.6088	0.0839	0.1428	0.5577	0.1255	0.1578	0.4281
VZV: progression to high titres	-0.2867	0.2766	0.3022	-0.0342	0.2151	0.8738	0.0154	0.2176	0.9435
VZV: persistently high titres	0.2059	0.1484	0.1669	0.0516	0.1182	0.6629	0.0230	0.1268	0.8562
Change in time since baseline coefficient in PLWH				-0.0040	0.0068	0.5567
CD4 (baseline)							-0.0002	0.0003	0.4480
CD4 (longitudinal)							-0.0009	0.0003	0.0015
log10 viral load (baseline)							0.0158	0.0388	0.6839

write.csv(tabAll,row.names=F,file=paste(sep="_",outPrefix,"PWV_regressionModels.csv"))

tabAll.pwv<-tabAll

tab4$term<-as.character(tab4$term)
tab3$term<-as.character(tab3$term)
allRows<-setdiff(unique(c(tab4$term,tab3$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,coefficient.unadj=NA,SE.unadj=NA,pvalue.unadj=NA,coefficient.adj=NA,SE.adj=NA,pvalue.adj=NA)

tabAll$coefficient.unadj<-tab4$estimate[match(tabAll$term,tab4$term)]
tabAll$SE.unadj<-tab4$std.error[match(tabAll$term,tab4$term)]
tabAll$pvalue.unadj<-tab4$p.value[match(tabAll$term,tab4$term)]

tabAll$coefficient.adj<-tab3$estimate[match(tabAll$term,tab3$term)]
tabAll$SE.adj<-tab3$std.error[match(tabAll$term,tab3$term)]
tabAll$pvalue.adj<-tab3$p.value[match(tabAll$term,tab3$term)]

rn<-c("Time since baseline visit (in months)","Sex: male","Age (in years)","Current smoker","Alcohol drinker","Waist-hip ratio","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","Existing CKD","CMV: progression to low titres","CMV: progression to high titres","CMV: persistently high titres","VZV: progression to low titres","VZV: progression to high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (longitudinal)","log10 viral load (baseline)")

options(knitr.kable.NA = '')

tabAll %>%
  dplyr::select(!term) %>%
  kableRn(digits=4,escape=F,row_names=rn,col.names=rep(c("Coefficient","SE","p"),2),caption="Model summaries for all models fitted to PWV as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only.") %>%
  add_header_above(header=c(" "=1,"Unadjusted models" = 3, "Multiple regression model" = 3)) %>%
  kable_styling(full_width=F)

Table 8.10: Model summaries for all models fitted to PWV as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only.
	Unadjusted models			Multiple regression model
	Coefficient	SE	p	Coefficient	SE	p
Time since baseline visit (in months)	-0.0258	0.0032	0.0000	-0.0139	0.0040	0.0007
Sex: male	0.1238	0.1127	0.2723	-0.0908	0.1192	0.4473
Age (in years)	0.0714	0.0075	0.0000	0.0603	0.0071	0.0000
Current smoker	-0.0461	0.1710	0.7874	0.0418	0.1662	0.8015
Alcohol drinker	0.0795	0.1254	0.5261	0.1280	0.1313	0.3300
Waist-hip ratio	1.4778	0.8260	0.0741	-0.7116	0.7293	0.3296
BMI	0.0044	0.0125	0.7215	-0.0096	0.0120	0.4233
Diastolic blood pressure	0.0472	0.0032	0.0000	0.0324	0.0063	0.0000
Systolic blood pressure	0.0307	0.0022	0.0000	0.0106	0.0041	0.0110
Glucose (baseline)	0.0727	0.0272	0.0077	0.0551	0.0232	0.0174
Cholesterol (baseline)	0.0716	0.0524	0.1724	-0.0011	0.0484	0.9813
Existing CKD	0.2933	0.2835	0.3025	0.0254	0.2329	0.9132
CMV: progression to low titres	-0.2952	0.3753	0.4319	-0.2147	0.3105	0.4894
CMV: progression to high titres	0.1934	0.1545	0.2111	0.0841	0.1431	0.5572
CMV: persistently high titres	0.1039	0.1370	0.4484	0.0306	0.1276	0.8106
VZV: progression to low titres	0.1312	0.1957	0.5046	0.1255	0.1578	0.4281
VZV: progression to high titres	-0.2537	0.2653	0.3402	0.0154	0.2176	0.9435
VZV: persistently high titres	0.1520	0.1568	0.3341	0.0230	0.1268	0.8562
CD4 (baseline)	-0.0002	0.0003	0.4890	-0.0002	0.0003	0.4480
CD4 (longitudinal)	-0.0011	0.0003	2e-04	-0.0009	0.0003	0.0015
log10 viral load (baseline)	0.0094	0.0386	0.8071	0.0158	0.0388	0.6839

tabAll.pwv.hiv<-tabAll

8.3 Combining CIMT and PWV on a single table

tabAll.cimt.hiv$term[tabAll.cimt.hiv$term=="physio.bp_dia_enrolment"]<-"physio.bp_dia"
tabAll.cimt.hiv$term[tabAll.cimt.hiv$term=="physio.bp_sys_enrolment"]<-"physio.bp_sys"
tabAll.cimt.hiv$term[tabAll.cimt.hiv$term=="lims.CD4_Absolute_changeFromBaselineAt24Months"]<-"lims.CD4_Absolute"

allRows<-unique(c("cimt.cimt_enrolment","visitNum",tabAll.cimt.hiv$term,tabAll.pwv.hiv$term))

tabAll<-data.frame(term=allRows,
                   cimt.coefficient.unadj=NA,cimt.SE.unadj=NA,cimt.pvalue.unadj=NA,cimt.coefficient.adj=NA,cimt.SE.adj=NA,cimt.pvalue.adj=NA,
                   pwv.coefficient.unadj=NA,pwv.SE.unadj=NA,pwv.pvalue.unadj=NA,pwv.coefficient.adj=NA,pwv.SE.adj=NA,pwv.pvalue.adj=NA)

tabAll$cimt.coefficient.unadj<-tabAll.cimt.hiv$coefficient.unadj[match(allRows,tabAll.cimt.hiv$term)]
tabAll$cimt.SE.unadj<-tabAll.cimt.hiv$SE.unadj[match(allRows,tabAll.cimt.hiv$term)]
tabAll$cimt.pvalue.unadj<-tabAll.cimt.hiv$pvalue.unadj[match(allRows,tabAll.cimt.hiv$term)]
tabAll$cimt.coefficient.adj<-tabAll.cimt.hiv$coefficient.adj[match(allRows,tabAll.cimt.hiv$term)]
tabAll$cimt.SE.adj<-tabAll.cimt.hiv$SE.adj[match(allRows,tabAll.cimt.hiv$term)]
tabAll$cimt.pvalue.adj<-tabAll.cimt.hiv$pvalue.adj[match(allRows,tabAll.cimt.hiv$term)]

tabAll$pwv.coefficient.unadj<-tabAll.pwv.hiv$coefficient.unadj[match(allRows,tabAll.pwv.hiv$term)]
tabAll$pwv.SE.unadj<-tabAll.pwv.hiv$SE.unadj[match(allRows,tabAll.pwv.hiv$term)]
tabAll$pwv.pvalue.unadj<-tabAll.pwv.hiv$pvalue.unadj[match(allRows,tabAll.pwv.hiv$term)]
tabAll$pwv.coefficient.adj<-tabAll.pwv.hiv$coefficient.adj[match(allRows,tabAll.pwv.hiv$term)]
tabAll$pwv.SE.adj<-tabAll.pwv.hiv$SE.adj[match(allRows,tabAll.pwv.hiv$term)]
tabAll$pwv.pvalue.adj<-tabAll.pwv.hiv$pvalue.adj[match(allRows,tabAll.pwv.hiv$term)]

options(knitr.kable.NA = '')

rn<-c("CIMT (baseline)","Time since baseline visit (in months)","Sex: male","Age (in years)","Current smoker","Alcohol drinker","Waist-hip ratio","BMI","Diastolic blood pressure (baseline) *","Systolic blood pressure (baseline) *","Glucose (baseline)","Cholesterol (baseline)","Existing CKD","CMV: progression to low titres","CMV: progression to high titres","CMV: persistently high titres","VZV: progression to low titres","VZV: progression to high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (change from baseline) *","log10 viral load (baseline)")

tabAll %>%
  dplyr::select(!term) %>%
  kableRn(digits=4,escape=F,row_names=rn,col.names=rep(c("Coefficient","SE","p"),4),caption="Model summaries for all models fitted to PWV as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only. Predictors indicated with an asterisks have longitudinal values for the PWV models and values as indicated in brackets for the CIMT models.") %>%
  add_header_above(header=c(" "=1,"Unadjusted models" = 3, "Multiple regression model" = 3,"Unadjusted models" = 3, "Multiple regression model" = 3)) %>%
  add_header_above(header=c(" "=1,"CIMT"=6,"PWV"=6)) %>%
  kable_styling(full_width=F)

Table 8.11: Model summaries for all models fitted to PWV as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only. Predictors indicated with an asterisks have longitudinal values for the PWV models and values as indicated in brackets for the CIMT models.
	CIMT						PWV
	Unadjusted models			Multiple regression model			Unadjusted models			Multiple regression model
	Coefficient	SE	p	Coefficient	SE	p	Coefficient	SE	p	Coefficient	SE	p
CIMT (baseline)	0.1034	0.0538	0.058	0.0558	0.0704	0.432
Time since baseline visit (in months)							-0.0258	0.0032	0.0000	-0.0139	0.0040	0.0007
Sex: male	0.0011	0.0073	0.876	-0.0044	0.0086	0.611	0.1238	0.1127	0.2723	-0.0908	0.1192	0.4473
Age (in years)	0.0011	0.0007	0.133	0.0010	0.0008	0.195	0.0714	0.0075	0.0000	0.0603	0.0071	0.0000
Current smoker	-0.0034	0.0103	0.742	-0.0071	0.0122	0.559	-0.0461	0.1710	0.7874	0.0418	0.1662	0.8015
Alcohol drinker	-0.0006	0.0085	0.946	0.0026	0.0099	0.795	0.0795	0.1254	0.5261	0.1280	0.1313	0.3300
Waist-hip ratio	-0.0311	0.0517	0.548	-0.0416	0.0547	0.447	1.4778	0.8260	0.0741	-0.7116	0.7293	0.3296
BMI	-0.0011	0.0009	0.222	-0.0006	0.0010	0.548	0.0044	0.0125	0.7215	-0.0096	0.0120	0.4233
Diastolic blood pressure (baseline) *	-0.0001	0.0003	0.693	-0.0005	0.0006	0.369	0.0472	0.0032	0.0000	0.0324	0.0063	0.0000
Systolic blood pressure (baseline) *	0.0000	0.0002	0.875	0.0004	0.0004	0.290	0.0307	0.0022	0.0000	0.0106	0.0041	0.0110
Glucose (baseline)	0.0009	0.0017	0.611	0.0011	0.0018	0.546	0.0727	0.0272	0.0077	0.0551	0.0232	0.0174
Cholesterol (baseline)	-0.0046	0.0036	0.205	-0.0044	0.0037	0.244	0.0716	0.0524	0.1724	-0.0011	0.0484	0.9813
Existing CKD	-0.0139	0.0165	0.401	-0.0184	0.0168	0.275	0.2933	0.2835	0.3025	0.0254	0.2329	0.9132
CMV: progression to low titres	-0.0082	0.0226	0.716	-0.0011	0.0234	0.961	-0.2952	0.3753	0.4319	-0.2147	0.3105	0.4894
CMV: progression to high titres	0.0174	0.0093	0.062	0.0144	0.0098	0.142	0.1934	0.1545	0.2111	0.0841	0.1431	0.5572
CMV: persistently high titres	0.0081	0.0097	0.403	0.0074	0.0101	0.461	0.1039	0.1370	0.4484	0.0306	0.1276	0.8106
VZV: progression to low titres	0.0222	0.0117	0.061	0.0219	0.0120	0.071	0.1312	0.1957	0.5046	0.1255	0.1578	0.4281
VZV: progression to high titres	0.0039	0.0156	0.801	0.0064	0.0158	0.684	-0.2537	0.2653	0.3402	0.0154	0.2176	0.9435
VZV: persistently high titres	0.0198	0.0092	0.032	0.0187	0.0093	0.046	0.1520	0.1568	0.3341	0.0230	0.1268	0.8562
CD4 (baseline)	0.0000	0.0000	0.668	0.0000	0.0000	0.592	-0.0002	0.0003	0.4890	-0.0002	0.0003	0.4480
CD4 (change from baseline) *	0.0000	0.0000	0.335	0.0000	0.0000	0.241	-0.0011	0.0003	2e-04	-0.0009	0.0003	0.0015
log10 viral load (baseline)	0.0011	0.0025	0.664	0.0001	0.0029	0.966	0.0094	0.0386	0.8071	0.0158	0.0388	0.6839

tabAll$pwv.pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$pwv.pvalue.adj)))
tabAll$pwv.pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",gsub(pattern="<span.*>2",replace="2",tabAll$pwv.pvalue.unadj))))
tabAll$cimt.pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt.pvalue.adj)))
tabAll$cimt.pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt.pvalue.unadj)))
write.csv(tabAll,row.names = F,file=paste(sep="_",outPrefix,"CIMT_PWV_PLWH_Table2.csv"))

8.4 Repeating this for all participants

tabAll.cimt$term[tabAll.cimt$term=="physio.bp_dia_enrolment"]<-"physio.bp_dia"
tabAll.cimt$term[tabAll.cimt$term=="physio.bp_sys_enrolment"]<-"physio.bp_sys"
tabAll.cimt$term[tabAll.cimt$term=="lims.CD4_Absolute_changeFromBaselineAt24Months"]<-"lims.CD4_Absolute"

allRows<-setdiff(unique(c("cimt.cimt_enrolment","visitNum",tabAll.cimt$term,tabAll.pwv$term)),c("lims.CD4_Absolute_enrolment","lims.CD4_Absolute","log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)"))

tabAll<-data.frame(term=allRows,
                   cimt.coefficient.unadj=NA,cimt.SE.unadj=NA,cimt.pvalue.unadj=NA,cimt.coefficient.adj=NA,cimt.SE.adj=NA,cimt.pvalue.adj=NA,
                   pwv.coefficient.unadj=NA,pwv.SE.unadj=NA,pwv.pvalue.unadj=NA,pwv.coefficient.adj=NA,pwv.SE.adj=NA,pwv.pvalue.adj=NA)

tabAll$cimt.coefficient.unadj<-tabAll.cimt$coefficient.unadj[match(allRows,tabAll.cimt$term)]
tabAll$cimt.SE.unadj<-tabAll.cimt$SE.unadj[match(allRows,tabAll.cimt$term)]
tabAll$cimt.pvalue.unadj<-tabAll.cimt$pvalue.unadj[match(allRows,tabAll.cimt$term)]
tabAll$cimt.coefficient.adj<-tabAll.cimt$coefficient.adj[match(allRows,tabAll.cimt$term)]
tabAll$cimt.SE.adj<-tabAll.cimt$SE.adj[match(allRows,tabAll.cimt$term)]
tabAll$cimt.pvalue.adj<-tabAll.cimt$pvalue.adj[match(allRows,tabAll.cimt$term)]

tabAll$pwv.coefficient.unadj<-tabAll.pwv$coefficient.unadj[match(allRows,tabAll.pwv$term)]
tabAll$pwv.SE.unadj<-tabAll.pwv$SE.unadj[match(allRows,tabAll.pwv$term)]
tabAll$pwv.pvalue.unadj<-tabAll.pwv$pvalue.unadj[match(allRows,tabAll.pwv$term)]
tabAll$pwv.coefficient.adj<-tabAll.pwv$coefficient.adj[match(allRows,tabAll.pwv$term)]
tabAll$pwv.SE.adj<-tabAll.pwv$SE.adj[match(allRows,tabAll.pwv$term)]
tabAll$pwv.pvalue.adj<-tabAll.pwv$pvalue.adj[match(allRows,tabAll.pwv$term)]

options(knitr.kable.NA = '')

rn<-c("CIMT (baseline)","Time since baseline visit (in months)","Participant type: PLWH","Sex: male","Age (in years)","Current smoker","Alcohol drinker","Waist-hip ratio","BMI","Diastolic blood pressure (baseline) *","Systolic blood pressure (baseline) *","Glucose (baseline)","Cholesterol (baseline)","Existing CKD","CMV: progression to low titres","CMV: progression to high titres","CMV: persistently high titres","VZV: progression to low titres","VZV: progression to high titres","VZV: persistently high titres","Interaction term: time since baseline x participant type")

tabAll %>%
  dplyr::select(!term) %>%
  kableRn(digits=4,escape=F,row_names=rn,col.names=rep(c("Coefficient","SE","p"),4),caption="Model summaries for all models fitted to PWV as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to data from all study participants. Predictors indicated with an asterisks have longitudinal values for the PWV models and values as indicated in brackets for the CIMT models.") %>%
  add_header_above(header=c(" "=1,"Unadjusted models" = 3, "Multiple regression model" = 3,"Unadjusted models" = 3, "Multiple regression model" = 3)) %>%
  add_header_above(header=c(" "=1,"CIMT"=6,"PWV"=6)) %>%
  kable_styling(full_width=F)

Table 8.12: Model summaries for all models fitted to PWV as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to data from all study participants. Predictors indicated with an asterisks have longitudinal values for the PWV models and values as indicated in brackets for the CIMT models.
	CIMT						PWV
	Unadjusted models			Multiple regression model			Unadjusted models			Multiple regression model
	Coefficient	SE	p	Coefficient	SE	p	Coefficient	SE	p	Coefficient	SE	p
CIMT (baseline)	0.0427	0.0431	0.326	0.0085	0.0558	0.880
Time since baseline visit (in months)							-0.0251	0.0027	0.0000	-0.0168	0.0059	0.0043
Participant type: PLWH	0.0164	0.0068	0.017	0.0098	0.0082	0.233	-0.2428	0.1212	0.0454	0.1003	0.1351	0.4578
Sex: male	0.0016	0.0059	0.790	-0.0032	0.0068	0.636	0.1025	0.1015	0.3130	-0.0523	0.1052	0.6198
Age (in years)	0.0008	0.0006	0.178	0.0008	0.0006	0.179	0.0778	0.0061	0.0000	0.0630	0.0058	0.0000
Current smoker	-0.0043	0.0087	0.620	-0.0068	0.0105	0.515	0.0728	0.1639	0.6573	0.0371	0.1561	0.8122
Alcohol drinker	0.0005	0.0073	0.942	0.0025	0.0088	0.774	0.0478	0.1225	0.6967	0.1466	0.1241	0.2383
Waist-hip ratio	-0.0387	0.0435	0.374	-0.0434	0.0461	0.347	2.0841	0.7452	0.0052	-0.7283	0.6481	0.2614
BMI	-0.0010	0.0006	0.092	-0.0001	0.0007	0.886	0.0101	0.0101	0.3180	-0.0100	0.0101	0.3211
Diastolic blood pressure (baseline) *	-0.0002	0.0002	0.355	-0.0004	0.0005	0.476	0.0472	0.0029	0.0000	0.0280	0.0050	0.0000
Systolic blood pressure (baseline) *	-0.0001	0.0001	0.555	0.0002	0.0003	0.494	0.0317	0.0019	0.0000	0.0136	0.0034	0.0001
Glucose (baseline)	0.0002	0.0014	0.898	0.0005	0.0015	0.739	0.0716	0.0257	0.0054	0.0493	0.0212	0.0199
Cholesterol (baseline)	-0.0046	0.0027	0.086	-0.0022	0.0030	0.467	0.1132	0.0446	0.0115	0.0094	0.0428	0.8259
Existing CKD	-0.0087	0.0136	0.521	-0.0133	0.0138	0.339	0.3579	0.2586	0.1677	0.0255	0.2131	0.9047
CMV: progression to low titres	-0.0081	0.0202	0.687	-0.0069	0.0205	0.736	-0.2632	0.3680	0.4748	-0.0776	0.3017	0.7971
CMV: progression to high titres	0.0179	0.0082	0.030	0.0123	0.0084	0.143	0.0982	0.1594	0.5384	0.1211	0.1398	0.3883
CMV: persistently high titres	0.0114	0.0087	0.193	0.0071	0.0088	0.423	0.0093	0.1325	0.9439	0.0776	0.1150	0.5004
VZV: progression to low titres	0.0201	0.0097	0.041	0.0182	0.0097	0.062	0.0925	0.1802	0.6088	0.0839	0.1428	0.5577
VZV: progression to high titres	0.0049	0.0139	0.725	0.0052	0.0142	0.714	-0.2867	0.2766	0.3022	-0.0342	0.2151	0.8738
VZV: persistently high titres	0.0163	0.0078	0.037	0.0137	0.0079	0.083	0.2059	0.1484	0.1669	0.0516	0.1182	0.6629
Interaction term: time since baseline x participant type										-0.0040	0.0068	0.5567

tabAll$pwv.pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$pwv.pvalue.adj)))
tabAll$pwv.pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",gsub(pattern="<span.*>2",replace="2",tabAll$pwv.pvalue.unadj))))
tabAll$cimt.pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt.pvalue.adj)))
tabAll$cimt.pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt.pvalue.unadj)))
write.csv(tabAll,row.names = F,file=paste(sep="_",outPrefix,"CIMT_PWV_AllParticipants_TableS4.csv"))

8.5 Repeating this while including ART treatment

# CIMT
tab3<-sumFitandSum_cimt.hiv("+crf.gender+crf.age+crf.smoke+crf.alcohol+physio.waisthip_enrolment+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+eGFR_existingCKD+cmvStatus+vzvStatus+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)+crf.artRegimen.2levels")

tab4.baseline<-sumFitandSum_cimt.hiv("")
tab4.sex<-sumFitandSum_cimt.hiv("+crf.gender")
tab4.age<-sumFitandSum_cimt.hiv("+crf.age")
tab4.smoke<-sumFitandSum_cimt.hiv("+crf.smoke")
tab4.alcohol<-sumFitandSum_cimt.hiv("+crf.alcohol")
tab4.waisthip<-sumFitandSum_cimt.hiv("+physio.waisthip_enrolment")
tab4.bmi<-sumFitandSum_cimt.hiv("+physio.bmi_enrolment")
tab4.diabp<-sumFitandSum_cimt.hiv("+physio.bp_dia_enrolment")
tab4.sysbp<-sumFitandSum_cimt.hiv("+physio.bp_sys_enrolment")
tab4.glucose<-sumFitandSum_cimt.hiv("+lims.Glucose_enrolment")
tab4.cholesterol<-sumFitandSum_cimt.hiv("+lims.Cholesterol_enrolment")
tab4.ckd<-sumFitandSum_cimt.hiv("+eGFR_existingCKD")
tab4.cmv<-sumFitandSum_cimt.hiv("+cmvStatus")
tab4.vzv<-sumFitandSum_cimt.hiv("+vzvStatus")
tab4.cd4base<-sumFitandSum_cimt.hiv("+lims.CD4_Absolute_enrolment")
tab4.cd4change<-sumFitandSum_cimt.hiv("+lims.CD4_Absolute_changeFromBaselineAt24Months")
tab4.vl<-sumFitandSum_cimt.hiv("+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")
tab4.art<-sumFitandSum_cimt.hiv("+crf.artRegimen.2levels")

tab4<-rbind(tab4.baseline,
            tab4.sex,
            tab4.age,
            tab4.smoke,
            tab4.alcohol,
            tab4.waisthip,
            tab4.bmi,
            tab4.diabp,
            tab4.sysbp,
            tab4.glucose,
            tab4.cholesterol,
            tab4.ckd,
            tab4.cmv,
            tab4.vzv,
            tab4.cd4base,
            tab4.cd4change,
            tab4.vl,
            tab4.art
)


tab4$term<-as.character(tab4$term)
tab3$term<-as.character(tab3$term)
allRows<-setdiff(unique(c(tab4$term,tab3$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,coefficient.unadj=NA,SE.unadj=NA,pvalue.unadj=NA,coefficient.adj=NA,SE.adj=NA,pvalue.adj=NA)

tabAll$coefficient.unadj<-tab4$estimate[match(tabAll$term,tab4$term)]
tabAll$SE.unadj<-tab4$std.error[match(tabAll$term,tab4$term)]
tabAll$pvalue.unadj<-tab4$p.value[match(tabAll$term,tab4$term)]

tabAll$coefficient.adj<-tab3$estimate[match(tabAll$term,tab3$term)]
tabAll$SE.adj<-tab3$std.error[match(tabAll$term,tab3$term)]
tabAll$pvalue.adj<-tab3$p.value[match(tabAll$term,tab3$term)]

rn<-c("Baseline CIMT","Sex: male","Age (in years)","Current smoker","Alcohol drinker","Waist-hip ratio","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose","Cholesterol","Existing CKD","CMV: progression to low titres","CMV: progression to high titres","CMV: persistently high titres","VZV: progression to low titres","VZV: progression to high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (change from Baseline at 24 months)","log10 viral load (baseline)","ART: no change to dolutegravir")

options(knitr.kable.NA = '')

tabAll %>%
  dplyr::select(!term) %>%
  kableRn(digits=3,escape=F,row_names=rn,col.names=rep(c("Coefficient","SE","p"),2),caption="Model summaries for all models fitted to CIMT as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  add_header_above(header=c(" "=1,"Unadjusted models" = 3, "Multiple regression model" = 3)) %>%
  kable_styling(full_width=F)

Table 8.13: Model summaries for all models fitted to CIMT as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only.
	Unadjusted models			Multiple regression model
	Coefficient	SE	p	Coefficient	SE	p
Baseline CIMT	0.103	0.054	0.058	0.054	0.070	0.451
Sex: male	0.001	0.007	0.876	-0.005	0.009	0.544
Age (in years)	0.001	0.001	0.133	0.001	0.001	0.187
Current smoker	-0.003	0.010	0.742	-0.007	0.012	0.586
Alcohol drinker	-0.001	0.009	0.946	0.002	0.010	0.833
Waist-hip ratio	-0.031	0.052	0.548	-0.042	0.055	0.448
BMI	-0.001	0.001	0.222	-0.001	0.001	0.550
Diastolic blood pressure	0.000	0.000	0.693	-0.001	0.001	0.377
Systolic blood pressure	0.000	0.000	0.875	0.000	0.000	0.294
Glucose	0.001	0.002	0.611	0.001	0.002	0.557
Cholesterol	-0.005	0.004	0.205	-0.004	0.004	0.244
Existing CKD	-0.014	0.017	0.401	-0.018	0.017	0.289
CMV: progression to low titres	-0.008	0.023	0.716	-0.002	0.023	0.927
CMV: progression to high titres	0.017	0.009	0.062	0.015	0.010	0.135
CMV: persistently high titres	0.008	0.010	0.403	0.008	0.010	0.425
VZV: progression to low titres	0.022	0.012	0.061	0.021	0.012	0.076
VZV: progression to high titres	0.004	0.016	0.801	0.006	0.016	0.715
VZV: persistently high titres	0.020	0.009	0.032	0.019	0.009	0.047
CD4 (baseline)	0.000	0.000	0.668	0.000	0.000	0.588
CD4 (change from Baseline at 24 months)	0.000	0.000	0.335	0.000	0.000	0.253
log10 viral load (baseline)	0.001	0.003	0.664	0.000	0.003	0.988
ART: no change to dolutegravir	-0.005	0.007	0.444	-0.005	0.007	0.442

tabAll.cimt.hiv<-tabAll

Note that the intercept in the unadjusted model gives the average change in CIMT over 24 months:

sumFitandSum_cimt.hiv("-cimt.cimt_enrolment") %>%
  dplyr::select(!c(term,rownum,statistic,df)) %>%
  kableRn(digits=3,escape=F,row_names=c("(Intercept)"),col.names=c("Coefficient","SE","p"),caption="Model summary for the basic unadjusted model fitted to CIMT as outcome. Model fitted to PLWH only.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  kable_styling(full_width=F)

Table 8.14: Model summary for the basic unadjusted model fitted to CIMT as outcome. Model fitted to PLWH only.
	Coefficient	SE	p
(Intercept)	0.033	0.003	<0.001

We can also pull this out for the adjusted model:

tab3 %>%
  dplyr::filter(term=="(Intercept)") %>%
  dplyr::select(!c(term,rownum,statistic,df)) %>%
  kableRn(digits=3,escape=F,row_names=c("(Intercept)"),col.names=c("Coefficient","SE","p"),caption="Model summary (showing intercept only!) for the fully adjusted model fitted to CIMT as outcome. Model fitted to PLWH only but including ART treatment.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  kable_styling(full_width=F)

Table 8.15: Model summary (showing intercept only!) for the fully adjusted model fitted to CIMT as outcome. Model fitted to PLWH only but including ART treatment.
	Coefficient	SE	p
(Intercept)	0.002	0.057	0.972

# PWV
tab3<-sumFitandSum_pwv.hiv("visitNum+crf.gender+crf.age+crf.smoke+crf.alcohol+physio.waisthip_enrolment+physio.bmi_enrolment+physio.bp_dia+physio.bp_sys+lims.Glucose_enrolment+lims.Cholesterol_enrolment+eGFR_existingCKD+cmvStatus+vzvStatus+lims.CD4_Absolute_enrolment+lims.CD4_Absolute+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)+crf.artRegimen.2levels")

tab4.timeonly<-sumFitandSum_pwv.hiv("visitNum")
tab4.sex<-sumFitandSum_pwv.hiv("visitNum+crf.gender")
tab4.age<-sumFitandSum_pwv.hiv("visitNum+crf.age")
tab4.smoke<-sumFitandSum_pwv.hiv("visitNum+crf.smoke")
tab4.alcohol<-sumFitandSum_pwv.hiv("visitNum+crf.alcohol")
tab4.waisthip<-sumFitandSum_pwv.hiv("visitNum+physio.waisthip_enrolment")
tab4.bmi<-sumFitandSum_pwv.hiv("visitNum+physio.bmi_enrolment")
tab4.diabp<-sumFitandSum_pwv.hiv("visitNum+physio.bp_dia")
tab4.sysbp<-sumFitandSum_pwv.hiv("visitNum+physio.bp_sys")
tab4.glucose<-sumFitandSum_pwv.hiv("visitNum+lims.Glucose_enrolment")
tab4.cholesterol<-sumFitandSum_pwv.hiv("visitNum+lims.Cholesterol_enrolment")
tab4.ckd<-sumFitandSum_pwv.hiv("visitNum+eGFR_existingCKD")
tab4.cmv<-sumFitandSum_pwv.hiv("visitNum+cmvStatus")
tab4.vzv<-sumFitandSum_pwv.hiv("visitNum+vzvStatus")
tab4.cd4base<-sumFitandSum_pwv.hiv("visitNum+lims.CD4_Absolute_enrolment")
tab4.cd4change<-sumFitandSum_pwv.hiv("visitNum+lims.CD4_Absolute")
tab4.vl<-sumFitandSum_pwv.hiv("visitNum+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")
tab4.art<-sumFitandSum_pwv.hiv("crf.artRegimen.2levels")

tab4<-rbind(tab4.timeonly,
            tab4.sex,
            tab4.age,
            tab4.smoke,
            tab4.alcohol,
            tab4.waisthip,
            tab4.bmi,
            tab4.diabp,
            tab4.sysbp,
            tab4.glucose,
            tab4.cholesterol,
            tab4.ckd,
            tab4.cmv,
            tab4.vzv,
            tab4.cd4base,
            tab4.cd4change,
            tab4.vl,
            tab4.art
)

tab4$term<-as.character(tab4$term)
tab3$term<-as.character(tab3$term)
allRows<-setdiff(unique(c(tab4$term,tab3$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,coefficient.unadj=NA,SE.unadj=NA,pvalue.unadj=NA,coefficient.adj=NA,SE.adj=NA,pvalue.adj=NA)

tabAll$coefficient.unadj<-tab4$estimate[match(tabAll$term,tab4$term)]
tabAll$SE.unadj<-tab4$std.error[match(tabAll$term,tab4$term)]
tabAll$pvalue.unadj<-tab4$p.value[match(tabAll$term,tab4$term)]

tabAll$coefficient.adj<-tab3$estimate[match(tabAll$term,tab3$term)]
tabAll$SE.adj<-tab3$std.error[match(tabAll$term,tab3$term)]
tabAll$pvalue.adj<-tab3$p.value[match(tabAll$term,tab3$term)]

rn<-c("Time since baseline visit (in months)","Sex: male","Age (in years)","Current smoker","Alcohol drinker","Waist-hip ratio","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","Existing CKD","CMV: progression to low titres","CMV: progression to high titres","CMV: persistently high titres","VZV: progression to low titres","VZV: progression to high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (longitudinal)","log10 viral load (baseline)","ART: no change to dolutegravir")

options(knitr.kable.NA = '')

tabAll %>%
  dplyr::select(!term) %>%
  kableRn(digits=4,escape=F,row_names=rn,col.names=rep(c("Coefficient","SE","p"),2),caption="Model summaries for all models fitted to PWV as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only.") %>%
  add_header_above(header=c(" "=1,"Unadjusted models" = 3, "Multiple regression model" = 3)) %>%
  kable_styling(full_width=F)

Table 8.16: Model summaries for all models fitted to PWV as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only.
	Unadjusted models			Multiple regression model
	Coefficient	SE	p	Coefficient	SE	p
Time since baseline visit (in months)	-0.0258	0.0032	0.0000	-0.0139	0.0040	0.0007
Sex: male	0.1238	0.1127	0.2723	-0.0983	0.1201	0.4143
Age (in years)	0.0714	0.0075	0.0000	0.0603	0.0071	0.0000
Current smoker	-0.0461	0.1710	0.7874	0.0460	0.1661	0.7821
Alcohol drinker	0.0795	0.1254	0.5261	0.1234	0.1312	0.3473
Waist-hip ratio	1.4778	0.8260	0.0741	-0.7116	0.7297	0.3298
BMI	0.0044	0.0125	0.7215	-0.0097	0.0120	0.4205
Diastolic blood pressure	0.0472	0.0032	0.0000	0.0325	0.0063	0.0000
Systolic blood pressure	0.0307	0.0022	0.0000	0.0104	0.0041	0.0117
Glucose (baseline)	0.0727	0.0272	0.0077	0.0548	0.0232	0.0180
Cholesterol (baseline)	0.0716	0.0524	0.1724	-0.0013	0.0485	0.9780
Existing CKD	0.2933	0.2835	0.3025	0.0302	0.2333	0.8970
CMV: progression to low titres	-0.2952	0.3753	0.4319	-0.2240	0.3114	0.4721
CMV: progression to high titres	0.1934	0.1545	0.2111	0.0868	0.1433	0.5457
CMV: persistently high titres	0.1039	0.1370	0.4484	0.0369	0.1281	0.7735
VZV: progression to low titres	0.1312	0.1957	0.5046	0.1209	0.1589	0.4483
VZV: progression to high titres	-0.2537	0.2653	0.3402	0.0096	0.2181	0.9648
VZV: persistently high titres	0.1520	0.1568	0.3341	0.0221	0.1267	0.8618
CD4 (baseline)	-0.0002	0.0003	0.4890	-0.0002	0.0003	0.4541
CD4 (longitudinal)	-0.0011	0.0003	2e-04	-0.0009	0.0003	0.0014
log10 viral load (baseline)	0.0094	0.0386	0.8071	0.0150	0.0388	0.7000
ART: no change to dolutegravir	-0.0655	0.1054	0.5343	-0.0493	0.0899	0.5836

tabAll.pwv.hiv<-tabAll

# Combining CIMT and PWV
tabAll.cimt.hiv$term[tabAll.cimt.hiv$term=="physio.bp_dia_enrolment"]<-"physio.bp_dia"
tabAll.cimt.hiv$term[tabAll.cimt.hiv$term=="physio.bp_sys_enrolment"]<-"physio.bp_sys"
tabAll.cimt.hiv$term[tabAll.cimt.hiv$term=="lims.CD4_Absolute_changeFromBaselineAt24Months"]<-"lims.CD4_Absolute"

allRows<-unique(c("cimt.cimt_enrolment","visitNum",tabAll.cimt.hiv$term,tabAll.pwv.hiv$term))

tabAll<-data.frame(term=allRows,
                   cimt.coefficient.unadj=NA,cimt.SE.unadj=NA,cimt.pvalue.unadj=NA,cimt.coefficient.adj=NA,cimt.SE.adj=NA,cimt.pvalue.adj=NA,
                   pwv.coefficient.unadj=NA,pwv.SE.unadj=NA,pwv.pvalue.unadj=NA,pwv.coefficient.adj=NA,pwv.SE.adj=NA,pwv.pvalue.adj=NA)

tabAll$cimt.coefficient.unadj<-tabAll.cimt.hiv$coefficient.unadj[match(allRows,tabAll.cimt.hiv$term)]
tabAll$cimt.SE.unadj<-tabAll.cimt.hiv$SE.unadj[match(allRows,tabAll.cimt.hiv$term)]
tabAll$cimt.pvalue.unadj<-tabAll.cimt.hiv$pvalue.unadj[match(allRows,tabAll.cimt.hiv$term)]
tabAll$cimt.coefficient.adj<-tabAll.cimt.hiv$coefficient.adj[match(allRows,tabAll.cimt.hiv$term)]
tabAll$cimt.SE.adj<-tabAll.cimt.hiv$SE.adj[match(allRows,tabAll.cimt.hiv$term)]
tabAll$cimt.pvalue.adj<-tabAll.cimt.hiv$pvalue.adj[match(allRows,tabAll.cimt.hiv$term)]

tabAll$pwv.coefficient.unadj<-tabAll.pwv.hiv$coefficient.unadj[match(allRows,tabAll.pwv.hiv$term)]
tabAll$pwv.SE.unadj<-tabAll.pwv.hiv$SE.unadj[match(allRows,tabAll.pwv.hiv$term)]
tabAll$pwv.pvalue.unadj<-tabAll.pwv.hiv$pvalue.unadj[match(allRows,tabAll.pwv.hiv$term)]
tabAll$pwv.coefficient.adj<-tabAll.pwv.hiv$coefficient.adj[match(allRows,tabAll.pwv.hiv$term)]
tabAll$pwv.SE.adj<-tabAll.pwv.hiv$SE.adj[match(allRows,tabAll.pwv.hiv$term)]
tabAll$pwv.pvalue.adj<-tabAll.pwv.hiv$pvalue.adj[match(allRows,tabAll.pwv.hiv$term)]

options(knitr.kable.NA = '')

rn<-c("CIMT (baseline)","Time since baseline visit (in months)","Sex: male","Age (in years)","Current smoker","Alcohol drinker","Waist-hip ratio","BMI","Diastolic blood pressure (baseline) *","Systolic blood pressure (baseline) *","Glucose (baseline)","Cholesterol (baseline)","Existing CKD","CMV: progression to low titres","CMV: progression to high titres","CMV: persistently high titres","VZV: progression to low titres","VZV: progression to high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (change from baseline) *","log10 viral load (baseline)","ART: no change from dolutegravir")

tabAll %>%
  dplyr::select(!term) %>%
  kableRn(digits=4,escape=F,row_names=rn,col.names=rep(c("Coefficient","SE","p"),4),caption="Model summaries for all models fitted to PWV as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only. Predictors indicated with an asterisks have longitudinal values for the PWV models and values as indicated in brackets for the CIMT models.") %>%
  add_header_above(header=c(" "=1,"Unadjusted models" = 3, "Multiple regression model" = 3,"Unadjusted models" = 3, "Multiple regression model" = 3)) %>%
  add_header_above(header=c(" "=1,"CIMT"=6,"PWV"=6)) %>%
  kable_styling(full_width=F)

Table 8.16: Model summaries for all models fitted to PWV as outcome (unadjusted, single variable models and multiple regression model.). Models fitted to PLWH only. Predictors indicated with an asterisks have longitudinal values for the PWV models and values as indicated in brackets for the CIMT models.
	CIMT						PWV
	Unadjusted models			Multiple regression model			Unadjusted models			Multiple regression model
	Coefficient	SE	p	Coefficient	SE	p	Coefficient	SE	p	Coefficient	SE	p
CIMT (baseline)	0.1034	0.0538	0.058	0.0535	0.0705	0.451
Time since baseline visit (in months)							-0.0258	0.0032	0.0000	-0.0139	0.0040	0.0007
Sex: male	0.0011	0.0073	0.876	-0.0053	0.0087	0.544	0.1238	0.1127	0.2723	-0.0983	0.1201	0.4143
Age (in years)	0.0011	0.0007	0.133	0.0010	0.0008	0.187	0.0714	0.0075	0.0000	0.0603	0.0071	0.0000
Current smoker	-0.0034	0.0103	0.742	-0.0067	0.0122	0.586	-0.0461	0.1710	0.7874	0.0460	0.1661	0.7821
Alcohol drinker	-0.0006	0.0085	0.946	0.0021	0.0099	0.833	0.0795	0.1254	0.5261	0.1234	0.1312	0.3473
Waist-hip ratio	-0.0311	0.0517	0.548	-0.0416	0.0547	0.448	1.4778	0.8260	0.0741	-0.7116	0.7297	0.3298
BMI	-0.0011	0.0009	0.222	-0.0006	0.0010	0.550	0.0044	0.0125	0.7215	-0.0097	0.0120	0.4205
Diastolic blood pressure (baseline) *	-0.0001	0.0003	0.693	-0.0005	0.0006	0.377	0.0472	0.0032	0.0000	0.0325	0.0063	0.0000
Systolic blood pressure (baseline) *	0.0000	0.0002	0.875	0.0004	0.0004	0.294	0.0307	0.0022	0.0000	0.0104	0.0041	0.0117
Glucose (baseline)	0.0009	0.0017	0.611	0.0010	0.0018	0.557	0.0727	0.0272	0.0077	0.0548	0.0232	0.0180
Cholesterol (baseline)	-0.0046	0.0036	0.205	-0.0044	0.0037	0.244	0.0716	0.0524	0.1724	-0.0013	0.0485	0.9780
Existing CKD	-0.0139	0.0165	0.401	-0.0178	0.0168	0.289	0.2933	0.2835	0.3025	0.0302	0.2333	0.8970
CMV: progression to low titres	-0.0082	0.0226	0.716	-0.0022	0.0234	0.927	-0.2952	0.3753	0.4319	-0.2240	0.3114	0.4721
CMV: progression to high titres	0.0174	0.0093	0.062	0.0147	0.0098	0.135	0.1934	0.1545	0.2111	0.0868	0.1433	0.5457
CMV: persistently high titres	0.0081	0.0097	0.403	0.0081	0.0101	0.425	0.1039	0.1370	0.4484	0.0369	0.1281	0.7735
VZV: progression to low titres	0.0222	0.0117	0.061	0.0215	0.0120	0.076	0.1312	0.1957	0.5046	0.1209	0.1589	0.4483
VZV: progression to high titres	0.0039	0.0156	0.801	0.0058	0.0158	0.715	-0.2537	0.2653	0.3402	0.0096	0.2181	0.9648
VZV: persistently high titres	0.0198	0.0092	0.032	0.0186	0.0093	0.047	0.1520	0.1568	0.3341	0.0221	0.1267	0.8618
CD4 (baseline)	0.0000	0.0000	0.668	0.0000	0.0000	0.588	-0.0002	0.0003	0.4890	-0.0002	0.0003	0.4541
CD4 (change from baseline) *	0.0000	0.0000	0.335	0.0000	0.0000	0.253	-0.0011	0.0003	2e-04	-0.0009	0.0003	0.0014
log10 viral load (baseline)	0.0011	0.0025	0.664	0.0000	0.0029	0.988	0.0094	0.0386	0.8071	0.0150	0.0388	0.7000
ART: no change from dolutegravir	-0.0054	0.0070	0.444	-0.0054	0.0070	0.442	-0.0655	0.1054	0.5343	-0.0493	0.0899	0.5836

tabAll$pwv.pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$pwv.pvalue.adj)))
tabAll$pwv.pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",gsub(pattern="<span.*>2",replace="2",tabAll$pwv.pvalue.unadj))))
tabAll$cimt.pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt.pvalue.adj)))
tabAll$cimt.pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt.pvalue.unadj)))
write.csv(tabAll,row.names = F,file=paste(sep="_",outPrefix,"CIMT_PWV_PLWH_withART_TableS3.csv"))

9 Summary of clinical and sub-clinical events

To provide a bit of context for the CVD and systemic event sections to follow, in Table 9.1 we summarise clinical and sub-clinical events, overall and for specific strata.

clinEventsTmp<-data.frame(
  pid=character(0),
  ptype=character(0),
  artnaive=integer(0),
  artRegimen=character(0),
  gender=character(0),
  age_group=character(0),
  cmvStatus=character(0),
  vzvStatus=character(0),
  cmvStatus_2levels=character(0),
  vzvStatus_2levels=character(0),
  eGFR_existingCKD=character(0),
  category=character(0)
)

pidDeath<-unique(c(ltfuDat$pid[ltfuDat$reason=="DEATH"]),clinEventsDat$PID[clinEventsDat$DIAGNOSIS=="death"])

for(j in which(!is.na(rhiccaDatWide$clinEventCategoryConfirmed) & rhiccaDatWide$clinEventCategoryConfirmed!="")){
  cats<-strsplit(split=";",rhiccaDatWide$clinEventCategoryConfirmed[j])
  clinEventsTmp<-rbind(clinEventsTmp,data.frame(pid=rhiccaDatWide$pid[j],ptype=rhiccaDatWide$crf.PLWH[j],artnaive=rhiccaDatWide$crf.artnaive[j],artRegimen=rhiccaDatWide$crf.artRegimen.2levels[j],gender=rhiccaDatWide$crf.gender[j],age_group=rhiccaDatWide$crf.age_group[j],cmvStatus=rhiccaDatWide$cmvStatus[j],vzvStatus=rhiccaDatWide$vzvStatus[j],cmvStatus_2levels=rhiccaDatWide$cmvStatus_2levels[j],vzvStatus_2levels=rhiccaDatWide$vzvStatus_2levels[j],eGFR_existingCKD=rhiccaDatWide$eGFR_existingCKD[j],category=unlist(cats)))
}

clinEventsTmp<-clinEventsTmp[!is.na(clinEventsTmp$category) & clinEventsTmp$category!="NA",]

#tmp<-rhiccaDatWide %>% dplyr::select(pid,crf.PLWH,crf.artnaive,crf.artRegimen.2levels,crf.gender,crf.age_group,cmvStatus,vzvStatus)
# for(j in 1:ncol(tmp)){tmp[,j]<-as.character(tmp[,j])}

# for(j in 1:nrow(clinEventsTmp)){
#   clinEventsTmp<-rbind(clinEventsTmp,unlist(c(clinEventsTmp[j,1:(ncol(clinEventsTmp)-1)],"Any")))
# }

tmpDeath<-rhiccaDatWide %>%
      dplyr::filter(pid %in% pidDeath) %>%
      dplyr::select(pid,crf.PLWH,crf.artnaive,crf.artRegimen.2levels,crf.gender,crf.age_group,cmvStatus,vzvStatus,cmvStatus_2levels,vzvStatus_2levels,eGFR_existingCKD)
colnames(tmpDeath)<-c("pid","ptype","artnaive","artRegimen","gender","age_group","cmvStatus","vzvStatus","cmvStatus_2levels","vzvStatus_2levels","eGFR_existingCKD")

clinEventsTmp<-cbind(
  rbind(
    clinEventsTmp[,-ncol(clinEventsTmp)],
    tmpDeath
  ),
  c(clinEventsTmp[,ncol(clinEventsTmp)],rep("Death",length(pidDeath)))
)

tmpCimt<-rhiccaDatWide %>%
      dplyr::filter(!is.na(rhiccaDatWide$newOnsetCIMT) & rhiccaDatWide$newOnsetCIMT=="new CIMT vasculopathy") %>%
      dplyr::select(pid,crf.PLWH,crf.artnaive,crf.artRegimen.2levels,crf.gender,crf.age_group,cmvStatus,vzvStatus,cmvStatus_2levels,vzvStatus_2levels,eGFR_existingCKD)
colnames(tmpCimt)<-c("pid","ptype","artnaive","artRegimen","gender","age_group","cmvStatus","vzvStatus","cmvStatus_2levels","vzvStatus_2levels","eGFR_existingCKD")

tmpPwv<-rhiccaDatWide %>%
      dplyr::filter(!is.na(rhiccaDatWide$newOnsetPWV) & rhiccaDatWide$newOnsetPWV=="new PWV vasculopathy") %>%
      dplyr::select(pid,crf.PLWH,crf.artnaive,crf.artRegimen.2levels,crf.gender,crf.age_group,cmvStatus,vzvStatus,cmvStatus_2levels,vzvStatus_2levels,eGFR_existingCKD)
colnames(tmpPwv)<-c("pid","ptype","artnaive","artRegimen","gender","age_group","cmvStatus","vzvStatus","cmvStatus_2levels","vzvStatus_2levels","eGFR_existingCKD")

clinEventsTmp<-cbind(
  rbind(
    clinEventsTmp[,-ncol(clinEventsTmp)],
    tmpCimt,
    tmpPwv
  ),
  c(clinEventsTmp[,ncol(clinEventsTmp)],rep("CIMT",sum(!is.na(rhiccaDatWide$newOnsetCIMT) & rhiccaDatWide$newOnsetCIMT=="new CIMT vasculopathy")),rep("PWV",sum(!is.na(rhiccaDatWide$newOnsetPWV) & rhiccaDatWide$newOnsetPWV=="new PWV vasculopathy")))
)

# for(pid in pidDeath){
#   clinEventsTmp<-rbind(clinEventsTmp,c(pid,unlist(as.character(tmp[tmp$pid==pid,c("crf.PLWH","crf.artnaive","crf.artRegimen.2levels","crf.gender","crf.age_group","cmvStatus","vzvStatus")])),"Death"))
# }
# for(pid in rhiccaDatWide$pid[!is.na(rhiccaDatWide$newOnsetCIMT) & rhiccaDatWide$newOnsetCIMT=="new CIMT vasculopathy"]){
#   clinEventsTmp<-rbind(clinEventsTmp,c(pid,unlist(as.character(tmp[tmp$pid==pid,c("crf.PLWH","crf.artnaive","crf.artRegimen.2levels","crf.gender","crf.age_group","cmvStatus","vzvStatus")])),"CIMT"))
# }

colnames(clinEventsTmp)[ncol(clinEventsTmp)]<-"category"

clinEventsSystemic<-clinEventsTmp %>% dplyr::filter(category=="Systemic disease")

tt<-table(clinEventsTmp$ptype,clinEventsTmp$category)
tt<-tt[,c("Death","Cancer","Infection","Systemic disease","Undetermined","CVD","CIMT","PWV")]
tt<-cbind(rowSums(tt[,!(colnames(tt) %in% c("Death","CIMT","PWV"))]),tt)
colnames(tt)[1]<-"Any clinical event"
colnames(tt)[ncol(tt)-1]<-"CVD (sub-clinical, CIMT)"
colnames(tt)[ncol(tt)]<-"CVD (sub-clinical, PWV)"
tt<-tt[,c(2,1,3:ncol(tt))]

ttCount<-tt
for(i in 1:nrow(ttCount)){
  tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[i]])/(1000*365.25))
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),")")
  }
}
 
ttPtype<-ttCount
 
#kable(ttCount,caption="Clinical events in participants, stratified by whether they are PLWH or controls. The number in brackets is the rate of events per 1,000 person-years.") %>%
#  kable_styling(full_width = FALSE)

tt<-table(clinEventsTmp$gender,clinEventsTmp$category)
tt<-tt[,c("Death","Cancer","Infection","Systemic disease","Undetermined","CVD","CIMT","PWV")]
tt<-cbind(rowSums(tt[,!(colnames(tt) %in% c("Death","CIMT","PWV"))]),tt)
colnames(tt)[1]<-"Any clinical event"
colnames(tt)[ncol(tt)-1]<-"CVD (sub-clinical, CIMT)"
colnames(tt)[ncol(tt)]<-"CVD (sub-clinical, PWV)"
tt<-tt[,c(2,1,3:ncol(tt))]

ttCount<-tt
for(i in 1:nrow(ttCount)){
  tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.gender==rownames(tt)[i]])/(1000*365.25))
}

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),")")
  }
}

ttSex<-ttCount

# kable(ttCount,caption="Clinical events stratified by sex. The number in brackets is the rate of events per 1,000 person-years.") %>%
#   kable_styling(full_width = FALSE)

tt<-table(clinEventsTmp$age_group,clinEventsTmp$category)
tt<-tt[,c("Death","Cancer","Infection","Systemic disease","Undetermined","CVD","CIMT","PWV")]
tt<-cbind(rowSums(tt[,!(colnames(tt) %in% c("Death","CIMT","PWV"))]),tt)
colnames(tt)[1]<-"Any clinical event"
colnames(tt)[ncol(tt)-1]<-"CVD (sub-clinical, CIMT)"
colnames(tt)[ncol(tt)]<-"CVD (sub-clinical, PWV)"
tt<-tt[,c(2,1,3:ncol(tt))]

ttCount<-tt
for(i in 1:nrow(ttCount)){
  tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age_group==rownames(tt)[i]])/(1000*365.25))
}

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),")")
  }
}

# kable(ttCount[rownames(ttCount)!="[0,35)",],caption="Clinical events stratified by age group. (All participants at least 35 years of age). The number in brackets is the rate of events per 1,000 person-years.") %>%
#   kable_styling(full_width = FALSE)

ttAgeGrp<-ttCount[rownames(ttCount)!="[0,35)",]

tt<-table(clinEventsTmp$cmvStatus,clinEventsTmp$category)
tt<-tt[,c("Death","Cancer","Infection","Systemic disease","Undetermined","CVD","CIMT","PWV")]
tt<-cbind(rowSums(tt[,!(colnames(tt) %in% c("Death","CIMT","PWV"))]),tt)
colnames(tt)[1]<-"Any clinical event"
colnames(tt)[ncol(tt)-1]<-"CVD (sub-clinical, CIMT)"
colnames(tt)[ncol(tt)]<-"CVD (sub-clinical, PWV)"
tt<-tt[,c(2,1,3:ncol(tt))]

ttCount<-tt
for(i in 1:nrow(ttCount)){
  tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$cmvStatus) & rhiccaDatWide$cmvStatus==rownames(tt)[i]])/(1000*365.25))
}

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),")")
  }
}

ttCount[ttCount=="0 (NaN)"]<-"0 (0)"

# kable(ttCount,caption="Clinical events stratified by CMV status. The number in brackets is the rate of events per 1,000 person-years.") %>%
#   kable_styling(full_width = FALSE)

ttCmv<-ttCount

tt<-table(clinEventsTmp$vzvStatus,clinEventsTmp$category)
tt<-tt[,c("Death","Cancer","Infection","Systemic disease","Undetermined","CVD","CIMT","PWV")]
tt<-cbind(rowSums(tt[,!(colnames(tt) %in% c("Death","CIMT","PWV"))]),tt)
colnames(tt)[1]<-"Any clinical event"
colnames(tt)[ncol(tt)-1]<-"CVD (sub-clinical, CIMT)"
colnames(tt)[ncol(tt)]<-"CVD (sub-clinical, PWV)"
tt<-tt[,c(2,1,3:ncol(tt))]

ttCount<-tt
for(i in 1:nrow(ttCount)){
  tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus==rownames(tt)[i]])/(1000*365.25))
}

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),")")
  }
}

# kable(ttCount,caption="Clinical events stratified by VZV status. The number in brackets is the rate of events per 1,000 person-years.") %>%
#   kable_styling(full_width = FALSE)

ttVzv<-ttCount

clinEventsTmpART<-clinEventsTmp %>% dplyr::filter(ptype=="PLWH")

tt<-table(clinEventsTmpART$artRegimen,clinEventsTmpART$category)
tt<-tt[,c("Death","Cancer","Infection","Systemic disease","Undetermined","CVD","CIMT","PWV")]
tt<-cbind(rowSums(tt[,!(colnames(tt) %in% c("Death","CIMT","PWV"))]),tt)
colnames(tt)[1]<-"Any clinical event"
colnames(tt)[ncol(tt)-1]<-"CVD (sub-clinical, CIMT)"
colnames(tt)[ncol(tt)]<-"CVD (sub-clinical, PWV)"
tt<-tt[,c(2,1,3:ncol(tt))]

ttCount<-tt
for(i in 1:nrow(ttCount)){
  tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$crf.artRegimen.2levels) & rhiccaDatWide$crf.artRegimen.2levels==rownames(tt)[i]])/(1000*365.25))
}

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),")")
  }
}
  
# kable(ttCount,caption="Clinical events in PLWH, stratified by whether they are switched ART regimen to dolutegravir or not. The number in brackets is the rate of events per 1,000 person-years.") %>%
#   kable_styling(full_width = FALSE)

ttArt<-ttCount

tt<-rbind(
  ttPtype,
  ttSex,
  ttAgeGrp,
  ttCmv,
  ttVzv,
  ttArt
)

kable(tt,caption=" Clinical and sub-clinical events summarised by different participant characteristics. The numbers in brackets are the event rates per 1,000 person-years.") %>%
  kableExtra::kable_styling(full_width = FALSE) %>%
  kableExtra::pack_rows(start_row=1,end_row=2,group_label="Participant type") %>%
  kableExtra::pack_rows(start_row=3,end_row=4,group_label="Sex") %>%
  kableExtra::pack_rows(start_row=5,end_row=8,group_label="Age group") %>%
  kableExtra::pack_rows(start_row=9,end_row=12,group_label="CMV status") %>%
  kableExtra::pack_rows(start_row=13,end_row=16,group_label="VZV status") %>%
  kableExtra::pack_rows(start_row=17,end_row=18,group_label="PLWH only: ART regimen")

Table 9.1: Clinical and sub-clinical events summarised by different participant characteristics. The numbers in brackets are the event rates per 1,000 person-years.
	Death	Any clinical event	Cancer	Infection	Systemic disease	Undetermined	CVD	CVD (sub-clinical, CIMT)	CVD (sub-clinical, PWV)
Participant type
Control	0 (0.000)	34 (82.831)	0 (0.000)	0 (0.000)	32 (77.958)	1 (2.436)	1 (2.436)	2 (4.872)	10 (24.362)
PLWH	37 (28.390)	100 (76.730)	5 (3.837)	27 (20.717)	57 (43.736)	3 (2.302)	8 (6.138)	9 (6.906)	11 (8.440)
Sex
Female	12 (17.184)	67 (95.945)	2 (2.864)	11 (15.752)	50 (71.601)	2 (2.864)	2 (2.864)	5 (7.160)	6 (8.592)
Male	25 (24.620)	67 (65.982)	3 (2.954)	16 (15.757)	39 (38.408)	2 (1.970)	7 (6.894)	6 (5.909)	15 (14.772)
Age group
[35,45)	13 (12.609)	61 (59.167)	3 (2.910)	12 (11.639)	41 (39.768)	3 (2.910)	2 (1.940)	1 (0.970)	7 (6.790)
[45,55)	16 (32.569)	49 (99.743)	1 (2.036)	10 (20.356)	34 (69.209)	1 (2.036)	3 (6.107)	2 (4.071)	9 (18.320)
[55,65)	6 (43.622)	12 (87.245)	0 (0.000)	3 (21.811)	6 (43.622)	0 (0.000)	3 (21.811)	5 (36.352)	3 (21.811)
[65,Inf)	2 (37.068)	12 (222.408)	1 (18.534)	2 (37.068)	8 (148.272)	0 (0.000)	1 (18.534)	3 (55.602)	2 (37.068)
CMV status
persistently low titres	4 (5.484)	32 (43.872)	0 (0.000)	4 (5.484)	25 (34.275)	1 (1.371)	2 (2.742)	7 (9.597)	11 (15.081)
progression to low titres	0 (0.000)	1 (45.486)	0 (0.000)	0 (0.000)	1 (45.486)	0 (0.000)	0 (0.000)	0 (0.000)	0 (0.000)
progression to high titres	1 (6.740)	13 (87.625)	1 (6.740)	0 (0.000)	12 (80.885)	0 (0.000)	0 (0.000)	2 (13.481)	1 (6.740)
persistently high titres	2 (10.143)	16 (81.145)	1 (5.072)	2 (10.143)	13 (65.931)	0 (0.000)	0 (0.000)	1 (5.072)	3 (15.215)
VZV status
persistently low titres	6 (8.101)	37 (49.958)	2 (2.700)	4 (5.401)	29 (39.157)	1 (1.350)	1 (1.350)	4 (5.401)	8 (10.802)
progression to low titres	0 (0.000)	7 (57.226)	0 (0.000)	1 (8.175)	6 (49.051)	0 (0.000)	0 (0.000)	1 (8.175)	2 (16.350)
progression to high titres	0 (0.000)	2 (42.669)	0 (0.000)	0 (0.000)	2 (42.669)	0 (0.000)	0 (0.000)	0 (0.000)	1 (21.335)
persistently high titres	1 (5.339)	16 (85.422)	0 (0.000)	1 (5.339)	14 (74.745)	0 (0.000)	1 (5.339)	5 (26.694)	4 (21.356)
PLWH only: ART regimen
ART: Change to dolutegravir	3 (4.112)	37 (50.713)	1 (1.371)	4 (5.482)	29 (39.748)	1 (1.371)	2 (2.741)	5 (6.853)	8 (10.965)
ART: No change to dolutegravir	34 (59.268)	63 (109.819)	4 (6.973)	23 (40.093)	28 (48.809)	2 (3.486)	6 (10.459)	4 (6.973)	3 (5.229)

10 Incidence rate of clinical and subclinical events are ~2-folds higher in PLWH.

In unadjusted analyses, shown in Table 10.1, we show that CVD clinical and sub-clinical CIMT event rates are higher in PLWH (though we lack power to determine that these differences are statistically significant), in particular event rates are 1.8 times higher (13.04 events per 1,000 person-years in PLWH compared to 7.31 in controls) when we consider clinical and sub-clinical events together. In the same table, we also show that pVZV is statistically significantly associated with clinical and sub-clinical events.

# idxCIMT<-which(!is.na(rhiccaDatWide$newOnsetCIMT) & rhiccaDatWide$newOnsetCIMT=="new CIMT vasculopathy")
# pidsCIMT<-rhiccaDatWide$pid[idxCIMT]   
# 
# tmp<-cbind(rhiccaDatWide %>% dplyr::filter(pid %in% pidsCIMT) %>% dplyr::select(c(pid,crf.PLWH,crf.artnaive,crf.artRegimen.2levels,crf.gender,crf.age_group,cmvStatus,vzvStatus)),"CVD (sub-clinical, CIMT)")
# colnames(tmp)<-colnames(clinEventsTmp)
# clinEventsTmp<-rbind(clinEventsTmp,tmp)

clinEventsTmp<-clinEventsTmp %>%
  dplyr::filter(category %in% c("CVD","CIMT","PWV"))

tt<-table(clinEventsTmp$category)
ttt<-data.frame(
  CVD.clinical=tt["CVD"],
  CVD.CIMT=tt["CIMT"],
  CVD.PWV=tt["PWV"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
tt<-tt/sum(as.numeric(rhiccaDatWide$obsTime)/(1000*365.25))
ttLow=tt-qnorm(0.975)*sqrt(tt/sum(as.numeric(rhiccaDatWide$obsTime)/(1000*365.25)))
ttUpp=tt+qnorm(0.975)*sqrt(tt/sum(as.numeric(rhiccaDatWide$obsTime)/(1000*365.25)))

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttLow[i,ttLow[i,]<0]<-0
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
  }
}

ttCountAll<-ttCount

tt<-table(clinEventsTmp$ptype,clinEventsTmp$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"],
  CVD.PWV=tt[,"PWV"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[i]])/(1000*365.25))
  ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[i]])/(1000*365.25)))
  ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[i]])/(1000*365.25)))
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttLow[i,ttLow[i,]<0]<-0
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
  }
}

#ttTmp<-cbind(ttCountCount[,"CVD.clinical"],c(sum(rhiccaDatWide$crf.ptype==rownames(tt)[1]),sum(rhiccaDatWide$crf.ptype==rownames(tt)[2]))-ttCountCount[,"CVD.clinical"])
#cvdClinicalFishP<-fisher.test(ttTmp)$p.value
cvdClinicalPoisP<-poisson.test(x=ttCountCount[,"CVD.clinical"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[1]])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[2]])/(1000*365.25))))$p.value

totCases<-sum(ttCountCount[,"CVD.clinical"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="PLWH","CVD.clinical"])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH!="PLWH"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="PLWH","CVD.clinical"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
pafClinical<-(totCases-hypoCases)/totCases

#ttTmp<-cbind(ttCountCount[,"CVD.CIMT"],c(sum(rhiccaDatWide$crf.ptype==rownames(tt)[1]),sum(rhiccaDatWide$crf.ptype==rownames(tt)[2]))-ttCountCount[,"CVD.CIMT"])
#cvdCimtFishP<-fisher.test(ttTmp)$p.value
cvdCimtPoisP<-poisson.test(x=ttCountCount[,"CVD.CIMT"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[1]])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[2]])/(1000*365.25))))$p.value

totCases<-sum(ttCountCount[,"CVD.CIMT"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="PLWH","CVD.CIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH!="PLWH"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="PLWH","CVD.CIMT"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
pafCimt<-(totCases-hypoCases)/totCases

# ttTmp<-cbind(ttCountCount[,"CVD.PWV"],c(sum(rhiccaDatWide$crf.ptype==rownames(tt)[1]),sum(rhiccaDatWide$crf.ptype==rownames(tt)[2]))-ttCountCount[,"CVD.PWV"])
# cvdPwvFishP<-fisher.test(ttTmp)$p.value
cvdPwvPoisP<-poisson.test(x=ttCountCount[,"CVD.PWV"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[1]])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[2]])/(1000*365.25))))$p.value

totCases<-sum(ttCountCount[,"CVD.PWV"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="PLWH","CVD.PWV"])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH!="PLWH"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="PLWH","CVD.PWV"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
pafPwv<-(totCases-hypoCases)/totCases

# ttTmp<-cbind(ttCountCount[,"CVD.clinicalAndCIMT"],c(sum(rhiccaDatWide$crf.ptype==rownames(tt)[1]),sum(rhiccaDatWide$crf.ptype==rownames(tt)[2]))-ttCountCount[,"CVD.clinicalAndCIMT"])
# cvdClinicalCimtFishP<-fisher.test(ttTmp)$p.value
cvdClinicalCimtPoisP<-poisson.test(x=ttCountCount[,"CVD.clinicalAndCIMT"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[1]])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[2]])/(1000*365.25))))$p.value

totCases<-sum(ttCountCount[,"CVD.clinicalAndCIMT"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="PLWH","CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH!="PLWH"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="PLWH","CVD.clinicalAndCIMT"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
pafClinCimt<-(totCases-hypoCases)/totCases

# ttTmp<-cbind(ttCountCount[,"CVD.clinicalAndCIMTAndPWV"],c(sum(rhiccaDatWide$crf.ptype==rownames(tt)[1]),sum(rhiccaDatWide$crf.ptype==rownames(tt)[2]))-ttCountCount[,"CVD.clinicalAndCIMTAndPWV"])
# cvdClinicalCimtPwvFishP<-fisher.test(ttTmp)$p.value
# cvdClinicalCimtPwvPoisP<-poisson.test(x=ttCountCount[,"CVD.clinicalAndCIMTAndPWV"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[1]])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==rownames(tt)[2]])/(1000*365.25))))$p.value

# kable(ttCount,caption="CVD clinical and sub-clinical events in participants, stratified by whether they are PLWH or controls. The number in brackets is the rate of events per 1,000 person-years.",col.names=paste(sep=", Poisson test p = ",c("Clinical events","CIMT sub-clinical events","PWV sub-clinical events","Clinical & CIMT sub-clinical events","Clinical & CIMT & PWV sub-clinical events"),format(nsmall=3,round(digits=3,c(cvdClinicalPoisP,cvdCimtPoisP,cvdPwvPoisP,cvdClinicalCimtPoisP,cvdClinicalCimtPwvPoisP))))) %>%
#   kable_styling(full_width = FALSE)

# rownames(ttCount)[rownames(ttCount)=="Patient"]<-"PLWH"

ttCountAll<-rbind(ttCountAll,paste(sep="","Poisson test p = ",format(nsmall=3,round(digits=3,c(cvdClinicalPoisP,cvdCimtPoisP,cvdPwvPoisP,cvdClinicalCimtPoisP))),", PAF = ",format(nsmall=3,round(digits=3,c(pafClinical,pafCimt,pafPwv,pafClinCimt)))),ttCount)

clinEventsTmp<-clinEventsTmp %>%
  dplyr::mutate(
    age_group_2lvls=case_when(
      age_group %in% c("[35,45)","[45,55)") ~ "<55",
      age_group %in% c("[55,65)","[65,Inf)") ~ "55+"
    )
  )

tt<-table(clinEventsTmp$age_group_2lvls,clinEventsTmp$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"],
  CVD.PWV=tt[,"PWV"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  if(i==1){
    tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55])/(1000*365.25))
    ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55])/(1000*365.25)))
    ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55])/(1000*365.25)))
  }else if(i==2){
    tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55])/(1000*365.25))
    ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55])/(1000*365.25)))
    ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55])/(1000*365.25)))
  }
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttLow[i,ttLow[i,]<0]<-0
    
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    
    if(rownames(ttCount)[i]=="<55"){
      airCvdUnder55=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdUnder55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdUnder55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdUnder55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
    }
    if(rownames(ttCount)[i]=="55+"){
      airCvdAbove55=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdAbove55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdAbove55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdAbove55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
    }
  }
}

cvdClinicalPoisP<-poisson.test(x=ttCountCount[,"CVD.clinical"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55])/(1000*365.25))))$p.value

totCases<-sum(ttCountCount[,"CVD.clinical"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.clinical"])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.clinical"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55])/(1000*365.25))
pafClinical<-(totCases-hypoCases)/totCases

cvdCimtPoisP<-poisson.test(x=ttCountCount[,"CVD.CIMT"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55])/(1000*365.25))))$p.value

totCases<-sum(ttCountCount[,"CVD.CIMT"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.CIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.CIMT"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55])/(1000*365.25))
pafCimt<-(totCases-hypoCases)/totCases

cvdPwvPoisP<-poisson.test(x=ttCountCount[,"CVD.PWV"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55])/(1000*365.25))))$p.value

totCases<-sum(ttCountCount[,"CVD.PWV"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.PWV"])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.PWV"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55])/(1000*365.25))
pafPwv<-(totCases-hypoCases)/totCases

cvdClinicalCimtPoisP<-poisson.test(x=ttCountCount[,"CVD.clinicalAndCIMT"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55])/(1000*365.25))))$p.value

totCases<-sum(ttCountCount[,"CVD.clinicalAndCIMT"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.clinicalAndCIMT"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55])/(1000*365.25))
pafClinCimt<-(totCases-hypoCases)/totCases

ttCountAll<-rbind(ttCountAll,paste(sep="","Poisson test p = ",format(nsmall=3,round(digits=3,c(cvdClinicalPoisP,cvdCimtPoisP,cvdPwvPoisP,cvdClinicalCimtPoisP))),", PAF = ",format(nsmall=3,round(digits=3,c(pafClinical,pafCimt,pafPwv,pafClinCimt)))),ttCount)

pidsWithVzv<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$vzvStatus)]
clinEventsTmp2<-clinEventsTmp %>%
  dplyr::filter(pid %in% pidsWithVzv) %>%
  dplyr::mutate(vzvStatus_2levels=ifelse(vzvStatus=="persistently high titres","persistently high titres","not persistently high titres"))

tt<-table(clinEventsTmp2$vzvStatus,clinEventsTmp2$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"],
  CVD.PWV=tt[,"PWV"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus==rownames(tt)[i]])/(1000*365.25))
    ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus==rownames(tt)[i]])/(1000*365.25)))
    ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus==rownames(tt)[i]])/(1000*365.25)))
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttLow[i,ttLow[i,]<0]<-0
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
  }
}

modClinEventsConfirmedCVD.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVD~vzvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsCIMTonly.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCIMTonly~vzvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsPWVonly.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantPWVonly~vzvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsConfirmedCVDwithCIMT.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMT~vzvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsConfirmedCVDwithCIMT.univariable.2levels<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMT~vzvStatus_2levels,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
# modClinEventsConfirmedCVDwithCIMTPWV.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMTPWV~vzvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))

cvdClinicalPoisP<-summary(pool(modClinEventsConfirmedCVD.univariable))$p.value[summary(pool(modClinEventsConfirmedCVD.univariable))$term=="vzvStatuspersistently high titres"]

totCases<-sum(ttCountCount[,"CVD.clinical"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinical"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus!="persistently high titres"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinical"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres"])/(1000*365.25))
pafClinical<-(totCases-hypoCases)/totCases

cvdCimtPoisP<-summary(pool(modClinEventsCIMTonly.univariable))$p.value[summary(pool(modClinEventsCIMTonly.univariable))$term=="vzvStatuspersistently high titres"]

totCases<-sum(ttCountCount[,"CVD.CIMT"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.CIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus!="persistently high titres"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.CIMT"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres"])/(1000*365.25))
pafCimt<-(totCases-hypoCases)/totCases

cvdPwvPoisP<-summary(pool(modClinEventsPWVonly.univariable))$p.value[summary(pool(modClinEventsPWVonly.univariable))$term=="vzvStatuspersistently high titres"]

totCases<-sum(ttCountCount[,"CVD.PWV"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.PWV"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus!="persistently high titres"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.PWV"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres"])/(1000*365.25))
pafPwv<-(totCases-hypoCases)/totCases

cvdClinicalCimtPoisP<-summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable))$p.value[summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable))$term=="vzvStatuspersistently high titres"]
cvdClinicalCimtPoisP.2levels<-summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable.2levels))$p.value[summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable.2levels))$term=="vzvStatus_2levelspersistently high titres"]

totCases<-sum(ttCountCount[,"CVD.clinicalAndCIMT"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus!="persistently high titres"])/(1000*365.25))
incExp<-sum(ttCountCount[rownames(ttCountCount)=="persistently high titres","CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres"])/(1000*365.25))
incOverall<-sum(ttCountCount[,"CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus)])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinicalAndCIMT"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres"])/(1000*365.25))
pafClinCimt<-(totCases-hypoCases)/totCases

# cvdClinicalCimtPwvPoisP<-summary(pool(modClinEventsConfirmedCVDwithCIMTPWV.univariable))$p.value[summary(pool(modClinEventsConfirmedCVDwithCIMTPWV.univariable))$term=="vzvStatuspersistently high titres"]

# cvdClinicalPoisP<-poisson.test(x=ttCountCount[,"CVD.clinical"],T=c(sum(!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$crf.ptype==rownames(tt)[1]),sum(!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$crf.ptype==rownames(tt)[2])))$p.value
# cvdCimtPoisP<-poisson.test(x=ttCountCount[,"CVD.CIMT"],T=c(sum(!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$crf.ptype==rownames(tt)[1]),sum(!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$crf.ptype==rownames(tt)[2])))$p.value
# cvdPwvPoisP<-poisson.test(x=ttCountCount[,"CVD.PWV"],T=c(sum(!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$crf.ptype==rownames(tt)[1]),sum(!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$crf.ptype==rownames(tt)[2])))$p.value
# cvdClinicalCimtPoisP<-poisson.test(x=ttCountCount[,"CVD.clinicalAndCIMT"],T=c(sum(!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$crf.ptype==rownames(tt)[1]),sum(!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$crf.ptype==rownames(tt)[2])))$p.value
# cvdClinicalCimtPwvPoisP<-poisson.test(x=ttCountCount[,"CVD.clinicalAndCIMTAndPWV"],T=c(!is.na(rhiccaDatWide$vzvStatus) & sum(rhiccaDatWide$crf.ptype==rownames(tt)[1]),sum(!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$crf.ptype==rownames(tt)[2])))$p.value

# kable(ttCount,caption="CVD clinical and sub-clinical events in participants, stratified by whether VZV status. The number in brackets is the rate of events per 1,000 person-years.",col.names=paste(sep=", Poisson regression p-value for persistently high titres = ",c("Clinical events","CIMT sub-clinical events","PWV sub-clinical events","Clinical & CIMT sub-clinical events","Clinical & CIMT & PWV sub-clinical events"),format(nsmall=3,round(digits=3,c(cvdClinicalPoisP,cvdCimtPoisP,cvdPwvPoisP,cvdClinicalCimtPoisP,cvdClinicalCimtPwvPoisP))))) %>%
#   kable_styling(full_width = FALSE)

ttCountAll<-rbind(ttCountAll,paste(sep="","Poisson regression p-value for persistently high VZV = ",format(nsmall=3,round(digits=3,c(cvdClinicalPoisP,cvdCimtPoisP,cvdPwvPoisP,cvdClinicalCimtPoisP))),", PAF = ",format(nsmall=3,round(digits=3,c(pafClinical,pafCimt,pafPwv,pafClinCimt)))),ttCount)

tt<-table(clinEventsTmp2$vzvStatus_2levels,clinEventsTmp2$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"],
  CVD.PWV=tt[,"PWV"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels==rownames(tt)[i]])/(1000*365.25))
    ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels==rownames(tt)[i]])/(1000*365.25)))
    ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels==rownames(tt)[i]])/(1000*365.25)))
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttLow[i,ttLow[i,]<0]<-0
    
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    if(rownames(ttCount)[i]=="not persistently high titres"){
      airCvdNotpVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdNotpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdNotpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdNotpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
    if(rownames(ttCount)[i]=="persistently high titres"){
      airCvdpVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
  }
}

pidsWithCmv<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$cmvStatus)]
clinEventsTmp2<-clinEventsTmp %>% dplyr::filter(pid %in% pidsWithCmv)

tt<-table(clinEventsTmp2$cmvStatus,clinEventsTmp2$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"],
  CVD.PWV=tt[,"PWV"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$cmvStatus) & rhiccaDatWide$cmvStatus==rownames(tt)[i]])/(1000*365.25))
  ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$cmvStatus) & rhiccaDatWide$cmvStatus==rownames(tt)[i]])/(1000*365.25)))
  ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$cmvStatus) & rhiccaDatWide$cmvStatus==rownames(tt)[i]])/(1000*365.25)))
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttLow[i,ttLow[i,]<0]<-0
    
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
  }
}

#ttCount<-ttCount[-which(grepl(pattern="NaN",ttCount[,1])),]

modClinEventsConfirmedCVD.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVD~cmvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsCIMTonly.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCIMTonly~cmvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsPWVonly.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantPWVonly~cmvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsConfirmedCVDwithCIMT.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMT~cmvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
# modClinEventsConfirmedCVDwithCIMTPWV.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMTPWV~cmvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))

cvdClinicalPoisP<-summary(pool(modClinEventsConfirmedCVD.univariable))$p.value[summary(pool(modClinEventsConfirmedCVD.univariable))$term=="cmvStatuspersistently high titres"]
cvdCimtPoisP<-summary(pool(modClinEventsCIMTonly.univariable))$p.value[summary(pool(modClinEventsCIMTonly.univariable))$term=="cmvStatuspersistently high titres"]
cvdPwvPoisP<-summary(pool(modClinEventsPWVonly.univariable))$p.value[summary(pool(modClinEventsPWVonly.univariable))$term=="cmvStatuspersistently high titres"]
cvdClinicalCimtPoisP<-summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable))$p.value[summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable))$term=="cmvStatuspersistently high titres"]
# cvdClinicalCimtPwvPoisP<-summary(pool(modClinEventsConfirmedCVDwithCIMTPWV.univariable))$p.value[summary(pool(modClinEventsConfirmedCVDwithCIMTPWV.univariable))$term=="cmvStatuspersistently high titres"]

# kable(ttCount,caption="CVD clinical and sub-clinical events in participants, stratified by whether CMV status. The number in brackets is the rate of events per 1,000 person-years.",col.names=paste(sep=", Poisson regression p-value for persistently high titres = ",c("Clinical events","CIMT sub-clinical events","PWV sub-clinical events","Clinical & CIMT sub-clinical events","Clinical & CIMT & PWV sub-clinical events"),format(nsmall=3,round(digits=3,c(cvdClinicalPoisP,cvdCimtPoisP,cvdPwvPoisP,cvdClinicalCimtPoisP,cvdClinicalCimtPwvPoisP))))) %>%
#   kable_styling(full_width = FALSE)

rownames(ttCount)<-paste(sep="",rownames(ttCount)," ")
ttCountAll<-rbind(ttCountAll,paste(sep="","Poisson regression p-value for persistently high CMV = ",format(nsmall=3,round(digits=3,c(cvdClinicalPoisP,cvdCimtPoisP,cvdPwvPoisP,cvdClinicalCimtPoisP)))),ttCount)

counter<-1
for(j in 1:nrow(ttCountAll)){
  if(grepl(rownames(ttCountAll)[j],pattern="[0-9]")){
    if(!(rownames(ttCountAll)[j] %in% c("<55","55+"))){
      rownames(ttCountAll)[j]<-paste(collapse="",rep(" ",counter-1))
      counter<-counter+1
    }
  }
}
rownames(ttCountAll)[1]<-paste(collapse="",rep(" ",counter-1))
rm(counter)

kable(ttCountAll,caption="CVD clinical and sub-clinical events in participants. The number in brackets is the rate of events per 1,000 person-years.",col.names=c("Clinical events","CIMT sub-clinical events","PWV sub-clinical events","Clinical & CIMT sub-clinical events")) %>%
  kable_styling(full_width = FALSE) %>%
  pack_rows(start_row=1,end_row=1,group_label="All participants.") %>%
  pack_rows(start_row=2,end_row=4,group_label="Stratified by participant type.") %>%
  pack_rows(start_row=5,end_row=7,group_label="Stratified by age group.") %>%
  pack_rows(start_row=8,end_row=12,group_label="Stratified by VZV titres.") %>%
  pack_rows(start_row=13,end_row=17,group_label="Stratified by CMV titres.")

Table 10.1: CVD clinical and sub-clinical events in participants. The number in brackets is the rate of events per 1,000 person-years.
	Clinical events	CIMT sub-clinical events	PWV sub-clinical events	Clinical & CIMT sub-clinical events
All participants.
	9 (5.252, 95% CI [1.821, 8.683])	11 (6.419, 95% CI [ 2.626, 10.212])	21 (12.254, 95% CI [ 7.013, 17.495])	20 (11.670, 95% CI [ 6.556, 16.785])
Stratified by participant type.
	Poisson test p = 0.696, PAF = 0.536	Poisson test p = 1.000, PAF = 0.241	Poisson test p = 0.018, PAF = -0.988	Poisson test p = 0.441, PAF = 0.374
Control	1 (2.436, 95% CI [0.000, 7.211])	2 (4.872, 95% CI [ 0.000, 11.625])	10 (24.362, 95% CI [ 9.263, 39.461])	3 (7.309, 95% CI [ 0.000, 15.579])
PLWH	8 (6.138, 95% CI [ 1.885, 10.392])	9 (6.906, 95% CI [ 2.394, 11.417])	11 (8.440, 95% CI [ 3.452, 13.428])	17 (13.044, 95% CI [ 6.843, 19.245])
Stratified by age group.
	Poisson test p = 0.012, PAF = 0.375	Poisson test p = 0.000, PAF = 0.693	Poisson test p = 0.077, PAF = 0.142	Poisson test p = 0.000, PAF = 0.550
<55	5 (3.285, 95% CI [0.406, 6.164])	3 (1.971, 95% CI [0.000, 4.201])	16 (10.511, 95% CI [ 5.361, 15.661])	8 (5.255, 95% CI [1.614, 8.897])
55+	4 (20.888, 95% CI [ 0.418, 41.358])	8 (41.776, 95% CI [12.827, 70.724])	5 (26.110, 95% CI [ 3.224, 48.996])	12 (62.664, 95% CI [27.209, 98.118])
Stratified by VZV titres.
	Poisson regression p-value for persistently high VZV = 0.827, PAF = 0.397	Poisson regression p-value for persistently high VZV = 0.007, PAF = 0.397	Poisson regression p-value for persistently high VZV = 0.334, PAF = 0.116	Poisson regression p-value for persistently high VZV = 0.025, PAF = 0.397
persistently low titres	1 (1.350, 95% CI [0.000, 3.997])	4 (5.401, 95% CI [ 0.108, 10.694])	8 (10.802, 95% CI [ 3.317, 18.287])	5 (6.751, 95% CI [ 0.834, 12.669])
progression to low titres	0 (0.000, 95% CI [0.000, 0.000])	1 (8.175, 95% CI [ 0.000, 24.198])	2 (16.350, 95% CI [ 0.000, 39.010])	1 (8.175, 95% CI [ 0.000, 24.198])
progression to high titres	0 (0.000, 95% CI [0.000, 0.000])	0 (0.000, 95% CI [0.000, 0.000])	1 (21.335, 95% CI [ 0.000, 63.150])	0 (0.000, 95% CI [0.000, 0.000])
persistently high titres	1 (5.339, 95% CI [ 0.000, 15.803])	5 (26.694, 95% CI [ 3.296, 50.093])	4 (21.356, 95% CI [ 0.427, 42.284])	6 (32.033, 95% CI [ 6.402, 57.665])
Stratified by CMV titres.
	Poisson regression p-value for persistently high CMV = 0.947	Poisson regression p-value for persistently high CMV = 0.974	Poisson regression p-value for persistently high CMV = 0.484	Poisson regression p-value for persistently high CMV = 0.927
persistently low titres	2 (2.742, 95% CI [0.000, 6.542])	7 (9.597, 95% CI [ 2.488, 16.707])	11 (15.081, 95% CI [ 6.169, 23.993])	9 (12.339, 95% CI [ 4.278, 20.400])
progression to low titres	0 (0.000, 95% CI [0.000, 0.000])	0 (0.000, 95% CI [0.000, 0.000])	0 (0.000, 95% CI [0.000, 0.000])	0 (0.000, 95% CI [0.000, 0.000])
progression to high titres	0 (0.000, 95% CI [0.000, 0.000])	2 (13.481, 95% CI [ 0.000, 32.164])	1 (6.740, 95% CI [ 0.000, 19.951])	2 (13.481, 95% CI [ 0.000, 32.164])
persistently high titres	0 (0.000, 95% CI [0.000, 0.000])	1 (5.072, 95% CI [ 0.000, 15.012])	3 (15.215, 95% CI [ 0.000, 32.432])	1 (5.072, 95% CI [ 0.000, 15.012])

tmp<-clinEventsTmp %>% dplyr::filter(!is.na(vzvStatus) & vzvStatus=="persistently high titres")
tt<-table(tmp$age_group_2lvls,tmp$category)
ttt<-data.frame(
  CVD.clinical=ifelse("CVD" %in% colnames(tt),tt[,"CVD"],0),
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  if(i==1){
    nEvents<-unlist(tt[i,])
    tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    
    tt[i,]<-nEvents/tObs
    ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
    ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
  }else if(i==2){
    nEvents<-unlist(tt[i,])
    tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    
    tt[i,]<-nEvents/tObs
    ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
    ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
  }
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    
    if(j==3){
      if(rownames(ttCount)[i]=="<55"){
        airCvdUnder55pVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdUnder55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdUnder55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdUnder55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
      if(rownames(ttCount)[i]=="55+"){
        airCvdAbove55pVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdAbove55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdAbove55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdAbove55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
    }
  }
}

tmp<-clinEventsTmp %>%
  dplyr::filter(!is.na(vzvStatus) & vzvStatus=="persistently high titres")

tt<-table(tmp$gender,tmp$category)
ttt<-data.frame(
  CVD.clinical=ifelse("CVD" %in% colnames(tt),tt[,"CVD"],0),
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
    nEvents<-unlist(tt[i,])
    tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.gender==rownames(ttCount)[i] & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    
    tt[i,]<-nEvents/tObs
    ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
    ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    
    if(j==3){
      if(rownames(ttCount)[i]=="Female"){
        airCvdFemalepVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdFemalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdFemalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdFemalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
      if(rownames(ttCount)[i]=="Male"){
        airCvdMalepVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdMalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdMalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdMalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
    }
  }
}

The absolute risk (expressed as an incidence rate) of clinical and sub-clinical CIMT events is 32.033, 95% CI [ 6.402, 57.665] per 1,000 person-years in participants with pVZV compared to 6.595, 95% CI [ 1.318, 11.872] per 1,000 person-years in those without pVZV (Poisson p-value for this difference p=0.026).

We can also express this in terms of 2-, 5- and 10-year risk. For this we make use of the cumulative distribution function of the Poisson distribution, using the above reported incidence rates and their associated 95% CIs. The Poisson test p-value for the difference in risk remains of course the same, regardless how the risk is reported.

The absolute risk (expressed as the probability of 1 or more events within the next 10 years; in other words the 10-year risk) of clinical and sub-clinical CIMT events is 27.41%, 95% CI [ 6.20%, 43.82%] in participants with pVZV compared to 6.38%, 95% CI [ 1.31%, 11.19%] in those without pVZV.

The absolute risk (expressed as the probability of 1 or more events within the next 5 years; in other words the 5-year risk) of clinical and sub-clinical CIMT events is 14.80%, 95% CI [ 3.15%, 25.05%] in participants with pVZV compared to 3.24%, 95% CI [0.66%, 5.76%] in those without pVZV.

The absolute risk (expressed as the probability of 1 or more events within the next 2 years; in other words the 2-year risk) of clinical and sub-clinical CIMT events is 6.21%, 95% CI [ 1.27%, 10.89%] in participants with pVZV compared to 1.31%, 95% CI [0.26%, 2.35%] in those without pVZV.

We see the association with pVZV also in minimally and fully adjusted models, though it is only borderline significant (Table 10.4).

The absolute risk (expressed as the probability of 1 or more events within the next 10 years; in other words the 10-year risk) of clinical and sub-clinical CIMT events is 46.56%, 95% CI [23.82%, 62.51%] in participants aged 55 and above compared to 5.12%, 95% CI [1.60%, 8.51%] in those below 55 years of age.

The absolute risk (expressed as the probability of 1 or more events within the next 5 years; in other words the 5-year risk) of clinical and sub-clinical CIMT events is 26.90%, 95% CI [12.72%, 38.77%] in participants aged 55 and above compared to 2.59%, 95% CI [0.80%, 4.35%] in those below 55 years of age.

The absolute risk (expressed as the probability of 1 or more events within the next 2 years; in other words the 2-year risk) of clinical and sub-clinical CIMT events is 11.78%, 95% CI [ 5.30%, 17.82%] in participants aged 55 and above compared to 1.05%, 95% CI [0.32%, 1.76%] in those below 55 years of age.

df<-data.frame(
  idx=1:6,
  pVZV=c("-",rep("+",5)),
  Grouping=factor(c(rep("All participants",2),rep("Age",2),rep("Sex",2)),levels=c("All participants","Age","Sex")),
  Characteristic=c(rep("",2),"<55","55+","Female","Male"),
  ir=c(as.numeric(strsplit(airCvdNotpVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdpVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdUnder55pVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdAbove55pVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdFemalepVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdMalepVZV,split=",")[[1]][1])),
  irLow=c(as.numeric(strsplit(strsplit(airCvdNotpVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdpVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdUnder55pVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdAbove55pVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdFemalepVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdMalepVZV,split=",")[[1]][2],split="\\[")[[1]][2])),
  irUpp=c(as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdNotpVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdpVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdUnder55pVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdAbove55pVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdFemalepVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdMalepVZV,split=",")[[1]][3]))),
  arFullCI=gsub(pattern="\\[ ",replacement="\\[",gsub(pattern=", 95% CI",replacement="",paste(sep="","AR ",c(
    ar2CvdNotpVZV,
    ar2CvdpVZV,
    ar2CvdUnder55pVZV,
    ar2CvdAbove55pVZV,
    ar2CvdFemalepVZV,
    ar2CvdMalepVZV
  ))))
) %>%
  dplyr::mutate(
    irFullCI=paste(sep="","IR ",format(nsmall=1,round(digits=1,ir))," [",format(nsmall=1,round(digits=1,irLow)),",",format(nsmall=1,round(digits=1,irUpp)),"]")
  )

gIrAll<-ggplot() +
  geom_bar(data=df,mapping=aes(x=idx,y=ir,fill=Grouping),stat="identity") +
  #geom_errorbar(data=df,mapping=aes(x=idx,y=ir,ymin=irLow,ymax=irUpp),width=0.1,alpha=0.75,lwd=0.4,col="darkgrey") +
  geom_text(data=df,mapping=aes(x=idx,y=ir+3.25,label=irFullCI),size=2.6,hjust=0.5,nudge_y = 2) +
  geom_text(data=df,mapping=aes(x=idx,y=ir,label=gsub(pattern=", ",replacement="",arFullCI)),size=2.6,hjust=0.5,nudge_y = 2) +
  geom_text(data=df,mapping=aes(label=pVZV,x=idx),y=-5,size=8) +
  geom_text(data=data.frame(lbl=c(rep("",2),"<55","55+","Female","Male"),x=1:6,col=df$Grouping),mapping=aes(label=lbl,x=x,col=col),y=-12,size=4) +
  geom_text(data=data.frame(lbl=factor(levels(df$Grouping),levels=levels(df$Grouping)),x=seq(1.5,5.5,by=2)),mapping=aes(label=lbl,x=x,col=lbl),y=-20,size=6) +
  geom_text(data=data.frame(lbl=c("-  no pVZV","+ pVZV"), x=c(1,1), y=c(105,110)), mapping=aes(x=x,y=y,label=lbl), size=5, hjust=0) +
  #geom_text(data=data.frame(x=4,y=130,lbl="//"),mapping=aes(x=x,y=y,label=lbl),col="darkgrey",size=5) +
  scale_fill_manual(values=c("steelblue","orange","#2E8B57")) +
  scale_color_manual(values=c("steelblue","orange","#2E8B57")) +
  coord_cartesian(ylim=c(-18,110)) +
  ylim(-25,1000) +
  theme_minimal() +
  theme(axis.text.x = element_blank(), axis.ticks.x = element_blank(),panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank(),text=element_text(size=16)) +
  guides(color=FALSE,fill=FALSE) +
  ylab("CBD/CVD sub-/clinical event rate (per 1,000 person-years)") +
  xlab("")

print(gIrAll)

pdf(width=8,height=8,file=paste(sep="_",outPrefix,"CIMT_CVDevents_FigureIncidenceRates_ALL_barplot.pdf"))
print(gIrAll)
dev.off()

## quartz_off_screen 
##                 2

We now repeat this for PLWH only.

clinEventsTmpPLWH<-clinEventsTmp %>%
  dplyr::filter(ptype=="PLWH")

tt<-table(clinEventsTmpPLWH$category)
ttt<-data.frame(
  CVD.clinical=tt["CVD"],
  CVD.CIMT=tt["CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt

nEvents<-unlist(tt)
tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))

tt<-nEvents/tObs
ttLow<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
ttUpp<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
# tt<-tt/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
# ttLow=tt-qnorm(0.975)*sqrt(tt/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
# ttUpp=tt+qnorm(0.975)*sqrt(tt/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))

tt<-t(as.matrix(tt))
ttLow<-t(as.matrix(ttLow))
ttUpp<-t(as.matrix(ttUpp))

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
  }
}

ttCountAll<-ttCount

tt<-table(clinEventsTmpPLWH$age_group_2lvls,clinEventsTmpPLWH$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  if(i==1){
    nEvents<-unlist(tt[i,])
    tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
    
    tt[i,]<-nEvents/tObs
    ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
    ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
    # tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
    # ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
    # ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
  }else if(i==2){
    nEvents<-unlist(tt[i,])
    tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))

    tt[i,]<-nEvents/tObs
    ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
    ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
    # tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
    # ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
    # ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
  }
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    
    if(rownames(ttCount)[i]=="<55"){
      airCvdUnder55=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdUnder55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdUnder55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdUnder55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
    }
    if(rownames(ttCount)[i]=="55+"){
      airCvdAbove55=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdAbove55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdAbove55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdAbove55=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
    }
  }
}

cvdClinicalPoisP<-poisson.test(x=ttCountCount[,"CVD.clinical"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))))$p.value

totCases<-sum(ttCountCount[,"CVD.clinical"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.clinical"])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.clinical"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
pafClinical<-(totCases-hypoCases)/totCases

cvdCimtPoisP<-poisson.test(x=ttCountCount[,"CVD.CIMT"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))))$p.value

totCases<-sum(ttCountCount[,"CVD.CIMT"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.CIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.CIMT"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
pafCimt<-(totCases-hypoCases)/totCases

cvdClinicalCimtPoisP<-poisson.test(x=ttCountCount[,"CVD.clinicalAndCIMT"],T=c(sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)),sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))))$p.value

totCases<-sum(ttCountCount[,"CVD.clinicalAndCIMT"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="55+","CVD.clinicalAndCIMT"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
pafClinCimt<-(totCases-hypoCases)/totCases

ttCountAll<-rbind(ttCountAll,paste(sep="","Poisson test p = ",format(nsmall=3,round(digits=3,c(cvdClinicalPoisP,cvdCimtPoisP,cvdClinicalCimtPoisP))),", PAF = ",format(nsmall=3,round(digits=3,c(pafClinical,pafCimt,pafClinCimt)))),ttCount)

tmp<-clinEventsTmpPLWH %>% dplyr::filter(!is.na(vzvStatus) & vzvStatus=="persistently high titres")
tt<-table(tmp$age_group_2lvls,tmp$category)
ttt<-data.frame(
  CVD.clinical=ifelse("CVD" %in% colnames(tt),tt[,"CVD"],0),
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  if(i==1){
    nEvents<-unlist(tt[i,])
    tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    
    tt[i,]<-nEvents/tObs
    ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
    ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
    # tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    # #nEvents<-
    # ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(log(tt[i,])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25)))
    # ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25)))
  }else if(i==2){
    nEvents<-unlist(tt[i,])
    tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    
    tt[i,]<-nEvents/tObs
    ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
    ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
    # tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    # ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25)))
    # ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25)))
  }
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    
    if(j==3){
      if(rownames(ttCount)[i]=="<55"){
        airCvdUnder55pVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdUnder55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdUnder55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdUnder55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
      if(rownames(ttCount)[i]=="55+"){
        airCvdAbove55pVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdAbove55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdAbove55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdAbove55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
    }
  }
}

tmp<-clinEventsTmpPLWH %>%
  dplyr::filter(!is.na(vzvStatus) & vzvStatus=="persistently high titres")

tt<-table(tmp$gender,tmp$category)
ttt<-data.frame(
  CVD.clinical=ifelse("CVD" %in% colnames(tt),tt[,"CVD"],0),
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
    nEvents<-unlist(tt[i,])
    tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.gender==rownames(ttCount)[i] & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    
    tt[i,]<-nEvents/tObs
    ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
    ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
    # tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    # #nEvents<-
    # ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(log(tt[i,])/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25)))
    # ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="PLWH" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25)))
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    
    if(j==3){
      if(rownames(ttCount)[i]=="Female"){
        airCvdFemalepVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdFemalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdFemalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdFemalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
      if(rownames(ttCount)[i]=="Male"){
        airCvdMalepVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdMalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdMalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdMalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
    }
  }
}

pidsWithVzv<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$vzvStatus)]
clinEventsTmpPLWH2<-clinEventsTmpPLWH %>%
  dplyr::filter(pid %in% pidsWithVzv) %>%
  dplyr::mutate(vzvStatus_2levels=ifelse(vzvStatus=="persistently high titres","persistently high titres","not persistently high titres"))

tt<-table(clinEventsTmpPLWH2$vzvStatus,clinEventsTmpPLWH2$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  nEvents<-unlist(tt[i,])
  tObs<-sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
  
  tt[i,]<-nEvents/tObs
  ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
  ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
  # tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
  #   ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
  #   ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
  }
}

modClinEventsConfirmedCVD.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(numClinEventsPerParticipantCVD~vzvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsCIMTonly.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(numClinEventsPerParticipantCIMTonly~vzvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsConfirmedCVDwithCIMT.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(numClinEventsPerParticipantCVDwithCIMT~vzvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsConfirmedCVDwithCIMT.univariable.2levels<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(numClinEventsPerParticipantCVDwithCIMT~vzvStatus_2levels,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))

cvdClinicalPoisP<-summary(pool(modClinEventsConfirmedCVD.univariable))$p.value[summary(pool(modClinEventsConfirmedCVD.univariable))$term=="vzvStatuspersistently high titres"]

totCases<-sum(ttCountCount[,"CVD.clinical"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinical"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus!="persistently high titres"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinical"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres"])/(1000*365.25))
pafClinical<-(totCases-hypoCases)/totCases

cvdCimtPoisP<-summary(pool(modClinEventsCIMTonly.univariable))$p.value[summary(pool(modClinEventsCIMTonly.univariable))$term=="vzvStatuspersistently high titres"]

totCases<-sum(ttCountCount[,"CVD.CIMT"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.CIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus!="persistently high titres" & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.CIMT"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres" & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
pafCimt<-(totCases-hypoCases)/totCases

# cvdPwvPoisP<-summary(pool(modClinEventsPWVonly.univariable))$p.value[summary(pool(modClinEventsPWVonly.univariable))$term=="vzvStatuspersistently high titres"]
cvdClinicalCimtPoisP<-summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable))$p.value[summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable))$term=="vzvStatuspersistently high titres"]
cvdClinicalCimtPoisP.2levels<-summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable.2levels))$p.value[summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable.2levels))$term=="vzvStatus_2levelspersistently high titres"]

totCases<-sum(ttCountCount[,"CVD.clinicalAndCIMT"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus!="persistently high titres" & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
incExp<-sum(ttCountCount[rownames(ttCountCount)=="persistently high titres","CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres" & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
incOverall<-sum(ttCountCount[,"CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinicalAndCIMT"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres" & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
pafClinCimt<-(totCases-hypoCases)/totCases
se<-sqrt(incNonExp/sum(rownames(ttCountCount)!="persistently high titres"))
incNonExpL<-incNonExp-qnorm(0.975)*se
incNonExpU<-incNonExp+qnorm(0.975)*se
hypoCasesL<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinicalAndCIMT"])+incNonExpL*sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres" & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
hypoCasesU<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinicalAndCIMT"])+incNonExpU*sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres" & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
pafClinCimtL<-(totCases-hypoCasesU)/totCases
pafClinCimtU<-(totCases-hypoCasesL)/totCases

ttCountAll<-rbind(ttCountAll,paste(sep="","Poisson regression p-value for persistently high VZV = ",format(nsmall=3,round(digits=3,c(cvdClinicalPoisP,cvdCimtPoisP,cvdClinicalCimtPoisP))),", PAF = ",format(nsmall=3,round(digits=3,c(pafClinical,pafCimt,pafClinCimt)))),ttCount)

tt<-table(clinEventsTmpPLWH2$vzvStatus_2levels,clinEventsTmpPLWH2$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  nEvents<-unlist(tt[i,])
  tObs<-sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
  
  tt[i,]<-nEvents/tObs
  ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
  ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
  # tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
  #   ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
  #   ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    if(rownames(ttCount)[i]=="not persistently high titres"){
      airCvdNotpVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdNotpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdNotpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdNotpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
    if(rownames(ttCount)[i]=="persistently high titres"){
      airCvdpVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
  }
}

tt<-table(clinEventsTmpPLWH2$gender,clinEventsTmpPLWH2$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  nEvents<-unlist(tt[i,])
  tObs<-sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$crf.gender) & rhiccaDatWide$crf.gender==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
  
  tt[i,]<-nEvents/tObs
  ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
  ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
  # tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$crf.gender) & rhiccaDatWide$crf.gender==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
  #   ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$crf.gender) & rhiccaDatWide$crf.gender==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
  #   ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$crf.gender) & rhiccaDatWide$crf.gender==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    if(rownames(ttCount)[i]=="Female"){
      airCvdFemale=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdFemale=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdFemale=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdFemale=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
    if(rownames(ttCount)[i]=="Male"){
      airCvdMale=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdMale=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdMale=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdMale=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
  }
}

pidsWithCmv<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$cmvStatus)]
clinEventsTmpPLWH2<-clinEventsTmpPLWH %>% dplyr::filter(pid %in% pidsWithCmv)

tt<-table(clinEventsTmpPLWH2$cmvStatus,clinEventsTmpPLWH2$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  nEvents<-unlist(tt[i,])
  tObs<-sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$cmvStatus) & rhiccaDatWide$cmvStatus==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
  
  tt[i,]<-nEvents/tObs
  ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
  ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
  # tt[i,]<-tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$cmvStatus) & rhiccaDatWide$cmvStatus==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
  # ttLow[i,]=tt[i,]-qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$cmvStatus) & rhiccaDatWide$cmvStatus==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
  # ttUpp[i,]=tt[i,]+qnorm(0.975)*sqrt(tt[i,]/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$cmvStatus) & rhiccaDatWide$cmvStatus==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25)))
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
  }
}

#ttCount<-ttCount[-which(grepl(pattern="NaN",ttCount[,1])),]

modClinEventsConfirmedCVD.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(numClinEventsPerParticipantCVD~cmvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsCIMTonly.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(numClinEventsPerParticipantCIMTonly~cmvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))
modClinEventsConfirmedCVDwithCIMT.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(numClinEventsPerParticipantCVDwithCIMT~cmvStatus,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))

cvdClinicalPoisP<-summary(pool(modClinEventsConfirmedCVD.univariable))$p.value[summary(pool(modClinEventsConfirmedCVD.univariable))$term=="cmvStatuspersistently high titres"]
cvdCimtPoisP<-summary(pool(modClinEventsCIMTonly.univariable))$p.value[summary(pool(modClinEventsCIMTonly.univariable))$term=="cmvStatuspersistently high titres"]
cvdClinicalCimtPoisP<-summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable))$p.value[summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable))$term=="cmvStatuspersistently high titres"]

rownames(ttCount)<-paste(sep="",rownames(ttCount)," ")
ttCountAll<-rbind(ttCountAll,paste(sep="","Poisson regression p-value for persistently high CMV = ",format(nsmall=3,round(digits=3,c(cvdClinicalPoisP,cvdCimtPoisP,cvdClinicalCimtPoisP)))),ttCount)

counter<-1
for(j in 1:nrow(ttCountAll)){
  if(grepl(rownames(ttCountAll)[j],pattern="[0-9]")){
    if(!(rownames(ttCountAll)[j] %in% c("<55","55+"))){
      rownames(ttCountAll)[j]<-paste(collapse="",rep(" ",counter-1))
      counter<-counter+1
    }
  }
}
#rownames(ttCountAll)[1]<-paste(collapse="",rep(" ",counter-1))
rm(counter)

kable(ttCountAll,caption="CVD clinical and sub-clinical events in PLWH. The number in brackets is the rate of events per 1,000 person-years.",col.names=c("Clinical events","CIMT sub-clinical events","Clinical & CIMT sub-clinical events")) %>%
  kable_styling(full_width = FALSE) %>%
  pack_rows(start_row=1,end_row=1,group_label="All participants.") %>%
  pack_rows(start_row=2,end_row=4,group_label="Stratified by age group.") %>%
  pack_rows(start_row=5,end_row=9,group_label="Stratified by VZV titres.") %>%
  pack_rows(start_row=10,end_row=14,group_label="Stratified by CMV titres.")

Table 10.2: CVD clinical and sub-clinical events in PLWH. The number in brackets is the rate of events per 1,000 person-years.
	Clinical events	CIMT sub-clinical events	Clinical & CIMT sub-clinical events
All participants.
CVD	8 (6.138, 95% CI [ 2.650, 12.095])	9 (6.906, 95% CI [ 3.158, 13.109])	17 (13.044, 95% CI [ 7.599, 20.885])
Stratified by age group.
	Poisson test p = 0.029, PAF = 0.313	Poisson test p = 0.000, PAF = 0.634	Poisson test p = 0.000, PAF = 0.483
<55	5 (4.217, 95% CI [1.369, 9.841])	3 (2.530, 95% CI [0.522, 7.394])	8 (6.747, 95% CI [ 2.913, 13.294])
55+	3 (25.519, 95% CI [ 5.263, 74.578])	6 (51.039, 95% CI [ 18.730, 111.090])	9 (76.558, 95% CI [ 35.007, 145.331])
Stratified by VZV titres.
	Poisson regression p-value for persistently high VZV = 0.996, PAF = -0.206	Poisson regression p-value for persistently high VZV = 0.004, PAF = 0.540	Poisson regression p-value for persistently high VZV = 0.054, PAF = 0.455
persistently low titres	1 (1.787, 95% CI [0.045, 9.957])	2 (3.574, 95% CI [ 0.433, 12.911])	3 (5.361, 95% CI [ 1.106, 15.667])
progression to low titres	0 (0.000, 95% CI [ 0.000, 35.734])	1 (9.687, 95% CI [ 0.245, 53.973])	1 (9.687, 95% CI [ 0.245, 53.973])
progression to high titres	0 (0.000, 95% CI [ 0.000, 82.377])	0 (0.000, 95% CI [ 0.000, 82.377])	0 (0.000, 95% CI [ 0.000, 82.377])
persistently high titres	0 (0.000, 95% CI [ 0.000, 22.989])	5 (31.160, 95% CI [10.118, 72.717])	5 (31.160, 95% CI [10.118, 72.717])
Stratified by CMV titres.
	Poisson regression p-value for persistently high CMV = 0.990	Poisson regression p-value for persistently high CMV = 0.938	Poisson regression p-value for persistently high CMV = 0.945
persistently low titres	1 (1.960, 95% CI [ 0.050, 10.922])	5 (9.801, 95% CI [ 3.182, 22.873])	6 (11.761, 95% CI [ 4.316, 25.600])
progression to low titres	0 (0.000, 95% CI [ 0.000, 167.791])	0 (0.000, 95% CI [ 0.000, 167.791])	0 (0.000, 95% CI [ 0.000, 167.791])
progression to high titres	0 (0.000, 95% CI [ 0.000, 26.102])	2 (14.152, 95% CI [ 1.714, 51.121])	2 (14.152, 95% CI [ 1.714, 51.121])
persistently high titres	0 (0.000, 95% CI [ 0.000, 19.218])	1 (5.210, 95% CI [ 0.132, 29.027])	1 (5.210, 95% CI [ 0.132, 29.027])

The absolute risk (expressed as an incidence rate) of clinical and sub-clinical CIMT events is 31.160, 95% CI [10.118, 72.717] per 1,000 person-years in PLWH with pVZV compared to 5.653, 95% CI [ 1.540, 14.474] per 1,000 person-years in those without pVZV (Poisson p-value for this difference p=0.057).

The absolute risk (expressed as the probability of 1 or more events within the next 10 years; in other words the 10-year risk) of clinical and sub-clinical CIMT events is 26.77%, 95% CI [ 9.62%, 51.67%] in PLWH with pVZV compared to 5.50%, 95% CI [ 1.53%, 13.47%] in those without pVZV.

The absolute risk (expressed as the probability of 1 or more events within the next 5 years; in other words the 5-year risk) of clinical and sub-clinical CIMT events is 14.43%, 95% CI [ 4.93%, 30.48%] in PLWH with pVZV compared to 2.79%, 95% CI [0.77%, 6.98%] in those without pVZV.

The absolute risk (expressed as the probability of 1 or more events within the next 2 years; in other words the 2-year risk) of clinical and sub-clinical CIMT events is 6.04%, 95% CI [ 2.00%, 13.54%] in PLWH with pVZV compared to 1.12%, 95% CI [0.31%, 2.85%] in those without pVZV.

The absolute risk (expressed as the probability of 1 or more events within the next 10 years; in other words the 10-year risk) of clinical and sub-clinical CIMT events is 53.49%, 95% CI [29.54%, 76.62%] in PLWH aged 55 and above compared to 6.52%, 95% CI [ 2.87%, 12.45%] in those below 55 years of age.

The absolute risk (expressed as the probability of 1 or more events within the next 5 years; in other words the 5-year risk) of clinical and sub-clinical CIMT events is 31.80%, 95% CI [16.06%, 51.65%] in PLWH aged 55 and above compared to 3.32%, 95% CI [1.45%, 6.43%] in those below 55 years of age.

The absolute risk (expressed as the probability of 1 or more events within the next 2 years; in other words the 2-year risk) of clinical and sub-clinical CIMT events is 14.20%, 95% CI [ 6.76%, 25.22%] in PLWH aged 55 and above compared to 1.34%, 95% CI [0.58%, 2.62%] in those below 55 years of age.

df<-data.frame(
  idx=1:6,
  pVZV=c("-",rep("+",5)),
  Grouping=factor(c(rep("All PLWH",2),rep("Age",2),rep("Sex",2)),levels=c("All PLWH","Age","Sex")),
  Characteristic=c(rep("",2),"<55","55+","Female","Male"),
  ir=c(as.numeric(strsplit(airCvdNotpVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdpVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdUnder55pVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdAbove55pVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdFemalepVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdMalepVZV,split=",")[[1]][1])),
  irLow=c(as.numeric(strsplit(strsplit(airCvdNotpVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdpVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdUnder55pVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdAbove55pVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdFemalepVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdMalepVZV,split=",")[[1]][2],split="\\[")[[1]][2])),
  irUpp=c(as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdNotpVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdpVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdUnder55pVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdAbove55pVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdFemalepVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdMalepVZV,split=",")[[1]][3]))),
  arFullCI=gsub(pattern="\\[ ",replacement="\\[",gsub(pattern=", 95% CI",replacement="",paste(sep="","AR ",c(
    ar2CvdNotpVZV,
    ar2CvdpVZV,
    ar2CvdUnder55pVZV,
    ar2CvdAbove55pVZV,
    ar2CvdFemalepVZV,
    ar2CvdMalepVZV
  ))))
) %>%
  dplyr::mutate(
    irFullCI=paste(sep="","IR ",format(nsmall=1,round(digits=1,ir))," [",format(nsmall=1,round(digits=1,irLow)),",",format(nsmall=1,round(digits=1,irUpp)),"]")
  )

gIrPlwh<-ggplot() +
  geom_bar(data=df,mapping=aes(x=idx,y=ir,fill=Grouping),stat="identity") +
  #geom_errorbar(data=df,mapping=aes(x=idx,y=ir,ymin=irLow,ymax=irUpp),width=0.1,alpha=0.75,lwd=0.4,col="darkgrey") +
  geom_text(data=df,mapping=aes(x=idx,y=ir+3.25,label=irFullCI),size=2.6,hjust=0.5,nudge_y = 2) +
  geom_text(data=df,mapping=aes(x=idx,y=ir,label=gsub(pattern=", ",replacement="",arFullCI)),size=2.6,hjust=0.5,nudge_y = 2) +
  geom_text(data=df,mapping=aes(label=pVZV,x=idx),y=-5,size=8) +
  geom_text(data=data.frame(lbl=c(rep("",2),"<55","55+","Female","Male"),x=1:6,col=df$Grouping),mapping=aes(label=lbl,x=x,col=col),y=-12,size=4) +
  geom_text(data=data.frame(lbl=factor(levels(df$Grouping),levels=levels(df$Grouping)),x=seq(1.5,5.5,by=2)),mapping=aes(label=lbl,x=x,col=lbl),y=-20,size=6) +
  geom_text(data=data.frame(lbl=c("-  no pVZV","+ pVZV"), x=c(1,1), y=c(105,110)), mapping=aes(x=x,y=y,label=lbl), size=5, hjust=0) +
  #geom_text(data=data.frame(x=4,y=130,lbl="//"),mapping=aes(x=x,y=y,label=lbl),col="darkgrey",size=5) +
  scale_fill_manual(values=c("steelblue","orange","#2E8B57")) +
  scale_color_manual(values=c("steelblue","orange","#2E8B57")) +
  coord_cartesian(ylim=c(-18,110)) +
  ylim(-25,1000) +
  theme_minimal() +
  theme(axis.text.x = element_blank(), axis.ticks.x = element_blank(),panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank(),text=element_text(size=16)) +
  guides(color=FALSE,fill=FALSE) +
  ylab("") +
  xlab("")

print(gIrPlwh + ylab("CBD/CVD sub-/clinical event rate (per 1,000 person-years)") )

pdf(width=8,height=8,file=paste(sep="_",outPrefix,"CIMT_CVDevents_FigureIncidenceRates_PLWH_barplot.pdf"))
print(gIrPlwh + ylab("CBD/CVD sub-/clinical event rate (per 1,000 person-years)") )
dev.off()

## quartz_off_screen 
##                 2

source("AbsRiskGraphs.R")

g_pVZV<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdpVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="orange",ttl="pVZV",cptn=paste(sep="","10-year CVD risk ",ar10CvdpVZV,"."))
g_NopVZV<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdNotpVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="darkgreen",ttl="No pVZV",cptn=paste(sep="","10-year CVD risk ",ar10CvdNotpVZV,"."))

g_Above55<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdAbove55pVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="orange",ttl="pVZV, age 55+ years",cptn=paste(sep="","10-year CVD risk ",ar10CvdAbove55pVZV,"."))
g_Under55<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdUnder55pVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="darkgreen",ttl="pVZV, age < 55 years",cptn=paste(sep="","10-year CVD risk ",ar10CvdUnder55pVZV,"."))

g_Male<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdMalepVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="orange",ttl="pVZV, male",cptn=paste(sep="","10-year CVD risk  ",ar10CvdMalepVZV,"."))

g_Female<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdFemalepVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="darkgreen",ttl="pVZV, female",cptn=paste(sep="","10-year CVD risk  ",ar10CvdFemalepVZV,"."))

grid.arrange(top="PLWH",g_pVZV,g_NopVZV,g_Above55,g_Under55,g_Male,g_Female,ncol=2)

pdf(width=5,height=9,file=paste(sep="_",outPrefix,"CIMT_CVDevents_FigureIncidenceRates_PLWH.pdf"))
grid.arrange(top="PLWH",g_pVZV,g_NopVZV,g_Above55,g_Under55,g_Male,g_Female,ncol=2)
dev.off()

And now we repeat the graph for controls.

clinEventsTmp$category<-factor(clinEventsTmp$category)
clinEventsTmp$age_group_2lvls<-factor(clinEventsTmp$age_group_2lvls)

clinEventsTmpCtrl<-clinEventsTmp %>%
  dplyr::filter(ptype=="Control")

tt<-table(clinEventsTmpCtrl$category)
ttt<-data.frame(
  CVD.clinical=tt["CVD"],
  CVD.CIMT=tt["CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt

nEvents<-unlist(tt)
tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="Control"])/(1000*365.25))

tt<-nEvents/tObs
ttLow<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
ttUpp<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs

tt<-t(as.matrix(tt))
ttLow<-t(as.matrix(ttLow))
ttUpp<-t(as.matrix(ttUpp))

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
  }
}


tmp<-clinEventsTmpCtrl %>% dplyr::filter(!is.na(vzvStatus) & vzvStatus=="persistently high titres")
tt<-table(tmp$age_group_2lvls,tmp$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  if(i==1){
    nEvents<-unlist(tt[i,])
    tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age<55 & rhiccaDatWide$crf.PLWH=="Control" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    
    tt[i,]<-nEvents/tObs
    ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
    ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
  }else if(i==2){
    nEvents<-unlist(tt[i,])
    tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.age>=55 & rhiccaDatWide$crf.PLWH=="Control" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    
    tt[i,]<-nEvents/tObs
    ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
    ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
  }
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    
    if(j==3){
      if(rownames(ttCount)[i]=="<55"){
        airCvdUnder55pVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdUnder55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdUnder55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdUnder55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
      if(rownames(ttCount)[i]=="55+"){
        airCvdAbove55pVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdAbove55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdAbove55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdAbove55pVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
    }
  }
}


tmp<-clinEventsTmpCtrl %>%
  dplyr::filter(!is.na(vzvStatus) & vzvStatus=="persistently high titres")

tt<-table(tmp$gender,tmp$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
    nEvents<-unlist(tt[i,])
    tObs<-sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.gender==rownames(ttCount)[i] & rhiccaDatWide$crf.PLWH=="Control" & !is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels=="persistently high titres"])/(1000*365.25))
    
    tt[i,]<-nEvents/tObs
    ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
    ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    
    if(j==3){
      if(rownames(ttCount)[i]=="Female"){
        airCvdFemalepVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdFemalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdFemalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdFemalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
      if(rownames(ttCount)[i]=="Male"){
        airCvdMalepVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
        ar10CvdMalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
        ar5CvdMalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
        ar2CvdMalepVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
    }
  }
}

pidsWithVzv<-rhiccaDatWide$pid[!is.na(rhiccaDatWide$vzvStatus)]
clinEventsTmpCtrl2<-clinEventsTmpCtrl %>%
  dplyr::filter(pid %in% pidsWithVzv) %>%
  dplyr::mutate(vzvStatus_2levels=ifelse(vzvStatus=="persistently high titres","persistently high titres","not persistently high titres"))

tt<-table(clinEventsTmpCtrl2$vzvStatus_2levels,clinEventsTmpCtrl2$category)
ttt<-data.frame(
  CVD.clinical=tt[,"CVD"],
  CVD.CIMT=tt[,"CIMT"]
) %>%
  dplyr::mutate(
    CVD.clinicalAndCIMT=CVD.clinical+CVD.CIMT
  )
tt<-ttt

ttCount<-tt
ttLow<-tt
ttUpp<-tt
for(i in 1:nrow(ttCount)){
  nEvents<-unlist(tt[i,])
  tObs<-sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus_2levels) & rhiccaDatWide$vzvStatus_2levels==rownames(tt)[i] & rhiccaDatWide$crf.PLWH=="Control"])/(1000*365.25))
  
  tt[i,]<-nEvents/tObs
  ttLow[i,]<-(qchisq(p=0.025,df=2*nEvents)/2)/tObs
  ttUpp[i,]<-(qchisq(p=0.975,df=2*(nEvents+1))/2)/tObs
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"])")
    if(rownames(ttCount)[i]=="not persistently high titres"){
      airCvdNotpVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdNotpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdNotpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdNotpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
    if(rownames(ttCount)[i]=="persistently high titres"){
      airCvdpVZV=paste(sep="",format(nsmall=3,round(digits=3,tt[i,j])),", 95% CI [",paste(collapse=", ",format(nsmall=3,round(digits=3,c(ttLow[i,j],ttUpp[i,j])))),"]")
      ar10CvdpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=10*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=10*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=10*ttUpp[i,j]/1000)))))),"%]")
      ar5CvdpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=5*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=5*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=5*ttUpp[i,j]/1000)))))),"%]")
      ar2CvdpVZV=paste(sep="",format(nsmall=2,round(digits=2,100*(1-ppois(q=0,lambda=2*tt[i,j]/1000)))),"%, 95% CI [",paste(collapse="%, ",format(nsmall=2,round(digits=2,c(100*(1-ppois(q=0,lambda=2*ttLow[i,j]/1000)),100*(1-ppois(q=0,lambda=2*ttUpp[i,j]/1000)))))),"%]")
      }
  }
}

totCases<-sum(ttCountCount[,"CVD.clinicalAndCIMT"])
incNonExp<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus!="persistently high titres" & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
incExp<-sum(ttCountCount[rownames(ttCountCount)=="persistently high titres","CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres" & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
incOverall<-sum(ttCountCount[,"CVD.clinicalAndCIMT"])/sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
hypoCases<-sum(ttCountCount[rownames(ttCountCount)!="persistently high titres","CVD.clinicalAndCIMT"])+incNonExp*sum(as.numeric(rhiccaDatWide$obsTime[!is.na(rhiccaDatWide$vzvStatus) & rhiccaDatWide$vzvStatus=="persistently high titres" & rhiccaDatWide$crf.PLWH=="PLWH"])/(1000*365.25))
pafClinCimt<-(totCases-hypoCases)/totCases


modClinEventsConfirmedCVDwithCIMT.univariable.2levels<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Control"),glm(numClinEventsPerParticipantCVDwithCIMT~vzvStatus_2levels,family="quasipoisson",offset=log(obsTime/(1000*365.25)+1)))

cvdClinicalCimtPoisP.2levels<-summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable.2levels))$p.value[summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable.2levels))$term=="vzvStatus_2levelspersistently high titres"]

df<-data.frame(
  idx=1:6,
  pVZV=c("-",rep("+",5)),
  Grouping=factor(c(rep("All Controls",2),rep("Age",2),rep("Sex",2)),levels=c("All Controls","Age","Sex")),
  Characteristic=c(rep("",2),"<55","55+","Female","Male"),
  ir=c(as.numeric(strsplit(airCvdNotpVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdpVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdUnder55pVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdAbove55pVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdFemalepVZV,split=",")[[1]][1]),
       as.numeric(strsplit(airCvdMalepVZV,split=",")[[1]][1])),
  irLow=c(as.numeric(strsplit(strsplit(airCvdNotpVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdpVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdUnder55pVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdAbove55pVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdFemalepVZV,split=",")[[1]][2],split="\\[")[[1]][2]),
          as.numeric(strsplit(strsplit(airCvdMalepVZV,split=",")[[1]][2],split="\\[")[[1]][2])),
  irUpp=c(as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdNotpVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdpVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdUnder55pVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdAbove55pVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdFemalepVZV,split=",")[[1]][3])),
          as.numeric(gsub(pattern=" |\\]",replace="",strsplit(airCvdMalepVZV,split=",")[[1]][3]))),
  arFullCI=gsub(pattern="\\[ ",replacement="\\[",gsub(pattern=", 95% CI",replacement="",paste(sep="","AR ",c(
    ar2CvdNotpVZV,
    ar2CvdpVZV,
    ar2CvdUnder55pVZV,
    ar2CvdAbove55pVZV,
    ar2CvdFemalepVZV,
    ar2CvdMalepVZV
  ))))
) %>%
  dplyr::mutate(
    irFullCI=paste(sep="","IR ",format(nsmall=1,round(digits=1,ir))," [",format(nsmall=1,round(digits=1,irLow)),",",format(nsmall=1,round(digits=1,irUpp)),"]")
  )

gIrCtrl<-ggplot() +
  geom_bar(data=df,mapping=aes(x=idx,y=ir,fill=Grouping),stat="identity") +
  #geom_errorbar(data=df,mapping=aes(x=idx,y=ir,ymin=irLow,ymax=irUpp),width=0.1,col="darkgrey",lwd=0.4,alpha=0.75) +
  geom_text(data=df,mapping=aes(x=idx,y=ir+3.25,label=irFullCI),size=2.6,hjust=0.5,nudge_y = 2) +
  geom_text(data=df,mapping=aes(x=idx,y=ir,label=gsub(pattern=", ",replacement="",arFullCI)),size=2.6,hjust=0.5,nudge_y = 2) +
  geom_text(data=df,mapping=aes(label=pVZV,x=idx),y=-5,size=8) +
  geom_text(data=data.frame(lbl=c(rep("",2),"<55","55+","Female","Male"),x=1:6,col=df$Grouping),mapping=aes(label=lbl,x=x,col=col),y=-12,size=4) +
  geom_text(data=data.frame(lbl=factor(levels(df$Grouping),levels=levels(df$Grouping)),x=seq(1.5,5.5,by=2)),mapping=aes(label=lbl,x=x,col=lbl),y=-20,size=6) +
  geom_text(data=data.frame(lbl=c("-  no pVZV","+ pVZV"), x=c(1,1), y=c(105,110)), mapping=aes(x=x,y=y,label=lbl), size=5, hjust=0) +
  #geom_text(data=data.frame(x=2,y=130,lbl="//"),mapping=aes(x=x,y=y,label=lbl),col="darkgrey",size=5) +
  #geom_text(data=data.frame(x=3,y=130,lbl="//"),mapping=aes(x=x,y=y,label=lbl),col="darkgrey",size=5) +
  #geom_text(data=data.frame(x=4,y=130,lbl="//"),mapping=aes(x=x,y=y,label=lbl),col="darkgrey",size=5) +
  #geom_text(data=data.frame(x=5,y=130,lbl="//"),mapping=aes(x=x,y=y,label=lbl),col="darkgrey",size=5) +
  #geom_text(data=data.frame(x=6,y=130,lbl="//"),mapping=aes(x=x,y=y,label=lbl),col="darkgrey",size=5) +
  scale_fill_manual(values=c("steelblue","orange","#2E8B57")) +
  scale_color_manual(values=c("steelblue","orange","#2E8B57")) +
  coord_cartesian(ylim=c(-18,110)) +
  ylim(-25,1000) +
  theme_minimal() +
  theme(axis.text.x = element_blank(), axis.ticks.x = element_blank(),panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank(),text=element_text(size=16)) +
  guides(color=FALSE,fill=FALSE) +
  ylab("") +
  xlab("")

print(gIrCtrl + ylab("CBD/CVD sub-/clinical event rate (per 1,000 person-years)"))

pdf(width=8,height=8,file=paste(sep="_",outPrefix,"CIMT_CVDevents_FigureIncidenceRates_Controls_barplot.pdf"))
print(gIrCtrl + ylab("CBD/CVD sub-/clinical event rate (per 1,000 person-years)"))
dev.off()

## quartz_off_screen 
##                 2

pdf(width=24,height=8,file=paste(sep="_",outPrefix,"SFig2_CIMT_CVDevents_FigureIncidenceRates_barplot.pdf"))
grid.arrange(
  gIrAll+labs(title="All participants") + theme(plot.title = element_text(hjust = 0.5)),
  gIrPlwh+labs(title="PLWH") + theme(plot.title = element_text(hjust = 0.5)),
  gIrCtrl+labs(title="HIV negative") + theme(plot.title = element_text(hjust = 0.5)),
  nrow=1)
dev.off()

## quartz_off_screen 
##                 2

pdf(width=8,height=8,file=paste(sep="_",outPrefix,"SFig2a_CIMT_CVDevents_FigureIncidenceRates_barplot_All.pdf"))
gIrAll+labs(title="All participants") + theme(plot.title = element_text(hjust = 0.5))
dev.off()

## quartz_off_screen 
##                 2

pdf(width=8,height=8,file=paste(sep="_",outPrefix,"SFig2b_CIMT_CVDevents_FigureIncidenceRates_barplot_PLWH.pdf"))
gIrPlwh+labs(title="PLWH") + theme(plot.title = element_text(hjust = 0.5)) + ylab("CBD/CVD sub-/clinical event rate (per 1,000 person-years)")
dev.off()

## quartz_off_screen 
##                 2

pdf(width=8,height=8,file=paste(sep="_",outPrefix,"SFig2c_CIMT_CVDevents_FigureIncidenceRates_barplot_Controls.pdf"))
gIrCtrl+labs(title="HIV negative") + theme(plot.title = element_text(hjust = 0.5)) + ylab("CBD/CVD sub-/clinical event rate (per 1,000 person-years)")
dev.off()

## quartz_off_screen 
##                 2

source("AbsRiskGraphs.R")

g_pVZV<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdpVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="orange",ttl="pVZV",cptn=paste(sep="","10-year CVD risk ",ar10CvdpVZV,"."))
g_NopVZV<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdNotpVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="darkgreen",ttl="No pVZV",cptn=paste(sep="","10-year CVD risk ",ar10CvdNotpVZV,"."))

g_Above55<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdAbove55pVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="orange",ttl="pVZV, age 55+ years",cptn=paste(sep="","10-year CVD risk ",ar10CvdAbove55pVZV,"."))
g_Under55<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdUnder55pVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="darkgreen",ttl="pVZV, age < 55 years",cptn=paste(sep="","10-year CVD risk ",ar10CvdUnder55pVZV,"."))

g_Male<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdMalepVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="orange",ttl="pVZV, male",cptn=paste(sep="","10-year CVD risk  ",ar10CvdMalepVZV,"."))
g_Female<-arPlot(ir=100*as.numeric(gsub(pattern="%",replacement="",strsplit(ar10CvdFemalepVZV,split=",")[[1]][1])),denom=10000,nrows=100,col="darkgreen",ttl="pVZV, female",cptn=paste(sep="","10-year CVD risk  ",ar10CvdFemalepVZV,"."))

grid.arrange(top="Controls",g_pVZV,g_NopVZV,g_Above55,g_Under55,g_Male,g_Female,ncol=2)

pdf(width=5,height=9,file=paste(sep="_",outPrefix,"CIMT_CVDevents_FigureIncidenceRates_Ctrl.pdf"))
grid.arrange(top="Controls",g_pVZV,g_NopVZV,g_Above55,g_Under55,g_Male,g_Female,ncol=2)
dev.off()

The absolute risk (expressed as an incidence rate) of clinical and sub-clinical CIMT events is 37.255, 95% CI [ 0.943, 207.573] per 1,000 person-years in controls with pVZV compared to 9.891, 95% CI [ 1.198, 35.730] per 1,000 person-years in those without pVZV (Poisson p-value for this difference p=0.250).

We can also express this in terms of 5-year and 10-year risk. For this we make use of the cumulative distribution function of the Poisson distribution, using the above reported incidence rates and their associated 95% CIs. The Poisson test p-value for the difference in risk remains of course the same, regardless how the risk is reported.

The absolute risk (expressed as the probability of 1 or more events within the next 10 years; in other words the 10-year risk) of clinical and sub-clinical CIMT events is 31.10%, 95% CI [ 0.94%, 87.45%] in controls with pVZV compared to 9.42%, 95% CI [ 1.19%, 30.04%] in those without pVZV.

In controls, the PAF for sub-clinical CIMT and clinical CVD events associated with pVZV is 0.182.

10.0.0.1 Extra figures for manuscript

PLWH - VZV, CMV and sub-/clinical events

gr<-expand.grid(c("No sub-/clinical CVD events","Sub-/clinical CVD events"),c("Enrolment","6months"),c("CMV","VZV"))

df<-data.frame(
  group=gr[,1],
  timepoint=gr[,2],
  marker=factor(gr[,3],levels=c("VZV","CMV")),
  n=NA,
  k=NA,
  prop=NA,
  propLow=NA,
  propUpp=NA
) %>%
  dplyr::mutate(
    timepointNum=case_when(timepoint=="Enrolment"~0,timepoint=="6months"~6),
    pValue=NA
  )

tmp<-clinEventsTmp #%>%
  #dplyr::filter(ptype=="PLWH")

pidsCvdEvent<-tmp$pid

tmp<-rhiccaDatWide %>%
  dplyr::filter(crf.ptype=="Patient") %>%
  dplyr::select(pid,vzv.vzv_enrolment,vzv.vzv_6months,cmv.cmv_enrolment,cmv.cmv_6months) %>%
  dplyr::mutate(
    group=factor(ifelse(pid %in% pidsCvdEvent,"Sub-/clinical CVD events","No sub-/clinical CVD events"))
  ) %>%
  tidyr::pivot_longer(cols=contains("_"),names_pattern="(vzv.vzv|cmv.cmv)_(.*)",names_to=c("marker","timepoint"),values_to="measurement") %>%
  dplyr::mutate(
    marker=factor(case_when(marker=="cmv.cmv"~"CMV",marker=="vzv.vzv"~"VZV"),levels=c("VZV","CMV")),
    timepoint=str_to_title(timepoint)
  )

for(i in 1:nrow(df)){
  df$n[i]<-sum(tmp$marker==df$marker[i] & tmp$group==df$group[i] & tmp$timepoint==df$timepoint[i] & !is.na(tmp$measurement))
  if(df$marker[i]=="VZV"){df$k[i]<-sum(tmp$marker==df$marker[i] & tmp$group==df$group[i] & tmp$timepoint==df$timepoint[i] & !is.na(tmp$measurement) & tmp$measurement>=vzvQuartiles["75%"])}
  if(df$marker[i]=="CMV"){df$k[i]<-sum(tmp$marker==df$marker[i] & tmp$group==df$group[i] & tmp$timepoint==df$timepoint[i] & !is.na(tmp$measurement) & tmp$measurement>=cmvQuartiles["75%"])}
  df$prop[i]<-df$k[i]/df$n[i]
  binomTmp<-binom.test(x=df$k[i],n=df$n[i])$conf.int
  df$propLow[i]<-binomTmp[1]
  df$propUpp[i]<-binomTmp[2]
  
  if(i %in% seq(2,nrow(df),by=2)){
    df$pValue[i-1]<-fisher.test(matrix(nrow=2,byrow=TRUE,c(df[i-1,"n"]-df[i-1,"k"],df[i-1,"k"],df[i,"n"]-df[i,"k"],df[i,"k"])))$p.value
  }
}

g1a<-df %>%
  dplyr::filter(marker=="VZV") %>%
  dplyr::mutate(
    timepointNum=case_when(group=="Sub-/clinical CVD events"~timepointNum-0.08,group=="No sub-/clinical CVD events"~timepointNum+0.08)
    ) %>%
  ggplot(mapping=aes(x=timepointNum,y=prop,ymin=propLow,ymax=propUpp,col=group,label=ifelse(pValue<0.05,"*","n.s."))) +
  geom_point(size=2) +
  geom_errorbar(width=0.2,alpha=0.65) +
  geom_line(lwd=0.85) +
  geom_text(mapping=aes(x=timepointNum,y=ifelse(!is.na(pValue) & pValue<0.05,0.845,0.85),size=ifelse(!is.na(pValue) & pValue<0.05,14,12),fontface=ifelse(!is.na(pValue) & pValue<0.05,"bold","plain")),col="black") +
  theme_minimal() +
  facet_wrap(~marker,nrow=1) +
  scale_color_manual(values=c("steelblue","orange"),name="") +
  scale_x_continuous(breaks=c(0,6),labels=c("Enrolment","6 months"),name="") +
  scale_size_continuous(limits = c(1,14)) +
  ylab("Proportion of IgG titres in top quartile") +
  guides(size=FALSE) +
  ylim(c(0,1)) +
  theme(axis.text.x = element_text(angle=45,hjust=1,vjust=1), panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank(),text=element_text(size=16),legend.position = "bottom",legend.direction="vertical")

g1b<-df %>%
  dplyr::filter(marker=="CMV") %>%
  dplyr::mutate(
    timepointNum=case_when(group=="Sub-/clinical CVD events"~timepointNum-0.08,group=="No sub-/clinical CVD events"~timepointNum+0.08)
    ) %>%
  ggplot(mapping=aes(x=timepointNum,y=prop,ymin=propLow,ymax=propUpp,col=group,label=ifelse(pValue<0.05,"*","n.s."))) +
  geom_point(size=2) +
  geom_errorbar(width=0.2,alpha=0.65) +
  geom_line(lwd=0.85) +
  geom_text(mapping=aes(x=timepointNum,y=ifelse(!is.na(pValue) & pValue<0.05,0.845,0.85),size=ifelse(!is.na(pValue) & pValue<0.05,14,12),fontface=ifelse(!is.na(pValue) & pValue<0.05,"bold","plain")),col="black") +
  theme_minimal() +
  facet_wrap(~marker,nrow=1) +
  scale_color_manual(values=c("steelblue","orange"),name="") +
  scale_x_continuous(breaks=c(0,6),labels=c("Enrolment","6 months"),name="") +
  scale_size_continuous(limits = c(1,14)) +
  ylab("") +
  guides(size=FALSE) +
  ylim(c(0,1)) +
  theme(axis.text.x = element_text(angle=45,hjust=1,vjust=1), panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank(),text=element_text(size=16),legend.position = "bottom",legend.direction="vertical")

dfRR<-data.frame(
  group=rep(c("not persistently high titres","persistently high titres"),2),
  RR=rep(c(1,NA),2),
  RRlogSE=NA,
  RRlow=NA,
  RRupp=NA,
  marker=factor(c(rep("VZV",2),rep("CMV",2)),levels=c("VZV","CMV")),
  sig=NA
)

# for(i in 1:nrow(dfRR)){
#   if(dfRR$marker[i]=="VZV"){
#     if(dfRR$group[i]!="persistently low titres"){
#       a0=sum(!is.na(clinEventsTmpPLWH$vzvStatus) & clinEventsTmpPLWH$vzvStatus=="persistently low titres")
#       a1=sum(!is.na(clinEventsTmpPLWH$vzvStatus) & clinEventsTmpPLWH$vzvStatus==dfRR$group[i])
#       t0=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.ptype=="Patient" & rhiccaDatWide$vzvStatus=="persistently low titres"],na.rm=TRUE))
#       t1=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.ptype=="Patient" & rhiccaDatWide$vzvStatus==dfRR$group[i]],na.rm=TRUE))
#       if((a0>0 | a1>0) & t0>0 & t1>0){
#         dfRR$RR[i]<-(a1/t1)/(a0/t0)
#         dfRR$RRlogSE[i]<-sqrt(1/a0+1/a1)
#         if(a1>0){
#           dfRR$RRlow[i]<-exp(log(dfRR$RR[i])-qnorm(0.975)*dfRR$RRlogSE[i])
#           dfRR$RRupp[i]<-exp(log(dfRR$RR[i])+qnorm(0.975)*dfRR$RRlogSE[i])
#         }else{
#           tmp<-rateratio.midp(x=c(a0,a1),y=c(t0,t1))
#           dfRR$RRlow[i]<-tmp$measure["Exposed2","lower"]
#           dfRR$RRupp[i]<-tmp$measure["Exposed2","upper"]
#         }
#         dfRR$sig[i]<-ifelse(dfRR$RRlow[i]<=1 & 1<=dfRR$RRupp[i],"n.s.","*")
#       }
#     }
#   }
#   if(dfRR$marker[i]=="CMV"){
#     if(dfRR$group[i]!="persistently low titres"){
#       a0=sum(!is.na(clinEventsTmpPLWH$cmvStatus) & clinEventsTmpPLWH$cmvStatus=="persistently low titres",dfRR$marker[i])
#       a1=sum(!is.na(clinEventsTmpPLWH$cmvStatus) & clinEventsTmpPLWH$cmvStatus==dfRR$group[i])
#       t0=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.ptype=="Patient" & rhiccaDatWide$cmvStatus=="persistently low titres"],na.rm=TRUE))
#       t1=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.ptype=="Patient" & rhiccaDatWide$cmvStatus==dfRR$group[i]],na.rm=TRUE))
#       if((a0>0 | a1>0) & t0>0 & t1>0){
#         dfRR$RR[i]<-(a1/t1)/(a0/t0)
#         dfRR$RRlogSE[i]<-sqrt(1/a0+1/a1)
#         if(a1>0){
#           dfRR$RRlow[i]<-exp(log(dfRR$RR[i])-qnorm(0.975)*dfRR$RRlogSE[i])
#           dfRR$RRupp[i]<-exp(log(dfRR$RR[i])+qnorm(0.975)*dfRR$RRlogSE[i])
#         }else{
#           tmp<-rateratio.midp(x=c(a0,a1),y=c(t0,t1))
#           dfRR$RRlow[i]<-tmp$measure["Exposed2","lower"]
#           dfRR$RRupp[i]<-tmp$measure["Exposed2","upper"]
#         }
#         dfRR$sig[i]<-ifelse(dfRR$RRlow[i]<=1 & 1<=dfRR$RRupp[i],"n.s.","*")
#       }
#     }
#   }
# }

for(i in 1:nrow(dfRR)){
  if(dfRR$marker[i]=="VZV"){
    if(dfRR$group[i]!="not persistently high titres"){
      a0=sum(clinEventsTmp$ptype=="PLWH" & !is.na(clinEventsTmp$vzvStatus_2levels) & clinEventsTmp$vzvStatus_2levels=="not persistently high titres")
      a1=sum(clinEventsTmp$ptype=="PLWH" & !is.na(clinEventsTmp$vzvStatus_2levels) & clinEventsTmp$vzvStatus_2levels==dfRR$group[i])
      t0=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH" & rhiccaDatWide$vzvStatus_2levels=="not persistently high titres"],na.rm=TRUE))
      t1=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH" & rhiccaDatWide$vzvStatus_2levels==dfRR$group[i]],na.rm=TRUE))
      if((a0>0 | a1>0) & t0>0 & t1>0){
        dfRR$RR[i]<-(a1/t1)/(a0/t0)
        dfRR$RRlogSE[i]<-sqrt(1/a0+1/a1)
        if(a1>0){
          dfRR$RRlow[i]<-exp(log(dfRR$RR[i])-qnorm(0.975)*dfRR$RRlogSE[i])
          dfRR$RRupp[i]<-exp(log(dfRR$RR[i])+qnorm(0.975)*dfRR$RRlogSE[i])
        }else{
          tmp<-rateratio.midp(x=c(a0,a1),y=c(t0,t1))
          dfRR$RRlow[i]<-tmp$measure["Exposed2","lower"]
          dfRR$RRupp[i]<-tmp$measure["Exposed2","upper"]
        }
        dfRR$sig[i]<-ifelse(dfRR$RRlow[i]<=1 & 1<=dfRR$RRupp[i],"n.s.","*")
      }
    }
  }
  if(dfRR$marker[i]=="CMV"){
    if(dfRR$group[i]!="not persistently high titres"){
      a0=sum(clinEventsTmp$ptype=="PLWH" & !is.na(clinEventsTmp$cmvStatus_2levels) & clinEventsTmp$cmvStatus_2levels=="not persistently high titres",dfRR$marker[i])
      a1=sum(clinEventsTmp$ptype=="PLWH" & !is.na(clinEventsTmp$cmvStatus_2levels) & clinEventsTmp$cmvStatus_2levels==dfRR$group[i])
      t0=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH" & rhiccaDatWide$cmvStatus_2levels=="not persistently high titres"],na.rm=TRUE))
      t1=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH=="PLWH" & rhiccaDatWide$cmvStatus_2levels==dfRR$group[i]],na.rm=TRUE))
      if((a0>0 | a1>0) & t0>0 & t1>0){
        dfRR$RR[i]<-(a1/t1)/(a0/t0)
        dfRR$RRlogSE[i]<-sqrt(1/a0+1/a1)
        if(a1>0){
          dfRR$RRlow[i]<-exp(log(dfRR$RR[i])-qnorm(0.975)*dfRR$RRlogSE[i])
          dfRR$RRupp[i]<-exp(log(dfRR$RR[i])+qnorm(0.975)*dfRR$RRlogSE[i])
        }else{
          tmp<-rateratio.midp(x=c(a0,a1),y=c(t0,t1))
          dfRR$RRlow[i]<-tmp$measure["Exposed2","lower"]
          dfRR$RRupp[i]<-tmp$measure["Exposed2","upper"]
        }
        dfRR$sig[i]<-ifelse(dfRR$RRlow[i]<=1 & 1<=dfRR$RRupp[i],"n.s.","*")
      }
    }
  }
}

g2a<-dfRR %>%
  dplyr::filter(marker=="VZV") %>%
  dplyr::mutate(
    group=factor(case_when(
      group=="not persistently high titres"~"Not persistently high titres\n(ref.)",
      group=="persistently high titres"~"Persistently high titres",
    ),levels=c("Not persistently high titres\n(ref.)","Persistently high titres")),
    RRupp=ifelse(is.nan(RRupp),500,dfRR$RRupp)
  ) %>%
  ggplot(mapping=aes(x=group,y=RR,ymin=RRlow,ymax=RRupp,label=sig)) +
  geom_point(size=3) +
  geom_errorbar(width=0.2,alpha=0.65) +
  geom_hline(yintercept=1,lty=2,col="darkgrey") +
  geom_text(mapping=aes(y=ifelse(sig=="*",32,35),size=ifelse(sig=="*",14,12),fontface=ifelse(sig=="*","bold","plain")),col="black") +
  theme_minimal() +
  facet_wrap(~marker,nrow=1) +
  #scale_color_manual(values=c("steelblue","orange"),name="") +
  #scale_x_continuous(breaks=c(0,6),labels=c("Enrolment","6 months"),name="") +
  ylab("Rate ratio for sub-/clinical events") +
  xlab("") +
  scale_size_continuous(limits = c(1,14)) +
  scale_y_continuous(trans="log10") +
  theme(axis.text.x = element_text(angle=45,hjust=1,vjust=1), panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank(),text=element_text(size=16),legend.position = "bottom") +
  guides(size=FALSE) +
  coord_cartesian(ylim=c(0.045,50))

g2b<-dfRR %>%
  dplyr::filter(marker=="CMV") %>%
  dplyr::mutate(
    group=factor(case_when(
      group=="not persistently high titres"~"Not persistently high titres\n(ref.)",
      group=="persistently high titres"~"Persistently high titres",
    ),levels=c("Not persistently high titres\n(ref.)","Persistently high titres")),
    RRupp=ifelse(is.nan(RRupp),500,dfRR$RRupp)
  ) %>%
  ggplot(mapping=aes(x=group,y=RR,ymin=RRlow,ymax=RRupp,label=sig)) +
  geom_point(size=3) +
  geom_errorbar(width=0.2,alpha=0.65) +
  geom_hline(yintercept=1,lty=2,col="darkgrey") +
  geom_text(mapping=aes(y=ifelse(sig=="*",32,35),size=ifelse(sig=="*",14,12),fontface=ifelse(sig=="*","bold","plain")),col="black") +
  theme_minimal() +
  facet_wrap(~marker,nrow=1) +
  #scale_color_manual(values=c("steelblue","orange"),name="") +
  #scale_x_continuous(breaks=c(0,6),labels=c("Enrolment","6 months"),name="") +
  ylab("") +
  xlab("") +
  scale_size_continuous(limits = c(1,14)) +
  scale_y_continuous(trans="log10") +
  theme(axis.text.x = element_text(angle=45,hjust=1,vjust=1), panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank(),text=element_text(size=16),legend.position = "bottom") +
  guides(size=FALSE) +
  coord_cartesian(ylim=c(0.045,50))

gPanelA<-(g1a | g1b) / (g2a | g2b) + plot_annotation(tag_levels=list(c("A","B","C","D")))

gTopA<-ggplot(mapping=aes(x=0.5,y=1,label="All participants")) + geom_text(size=16,fontface="bold") + theme_void()

pdf(width=7.5,height=12,file=paste(sep="_",outPrefix,"Fig2_HerpesGroups_HerpesTitres_RRofEvents_PLWH.pdf"))
(g1a | g1b) / (g2a | g2b) + plot_annotation(tag_levels=list(c("A","B","C","D")))
dev.off()

## quartz_off_screen 
##                 2

PLWH - CD4, VL and sub-/clinical events

gr<-expand.grid(c("No sub-/clinical CVD events","Sub-/clinical CVD events"),c("Enrolment","6months","12months"),c("CD4+ count","HIV VL"))

df<-data.frame(
  group=gr[,1],
  timepoint=gr[,2],
  marker=factor(gr[,3],levels=c("CD4+ count","HIV VL")),
  n=NA,
  k=NA,
  prop=NA,
  propLow=NA,
  propUpp=NA
) %>%
  dplyr::mutate(
    timepointNum=case_when(timepoint=="Enrolment"~0,timepoint=="6months"~6,timepoint=="12months"~12),
    pValue=NA
  )

tmp<-clinEventsTmp %>%
  dplyr::filter(ptype=="PLWH")

pidsCvdEvent<-tmp$pid

tmp<-rhiccaDatWide %>%
  dplyr::filter(crf.ptype=="Patient") %>%
  dplyr::select(pid,lims.CD4_Absolute_enrolment,lims.CD4_Absolute_6months,lims.CD4_Absolute_12months,lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment,lims.HIV_Abbott_Viral_Load_0pt2ml_P_6months,lims.HIV_Abbott_Viral_Load_0pt2ml_P_12months) %>%
  dplyr::mutate(
    group=factor(ifelse(pid %in% pidsCvdEvent,"Sub-/clinical CVD events","No sub-/clinical CVD events"))
  ) %>%
  tidyr::pivot_longer(cols=contains("_"),names_pattern="(lims.CD4_Absolute|lims.HIV_Abbott_Viral_Load_0pt2ml_P)_(.*)",names_to=c("marker","timepoint"),values_to="measurement") %>%
  dplyr::mutate(
    marker=factor(case_when(marker=="lims.CD4_Absolute"~"CD4+ count",marker=="lims.HIV_Abbott_Viral_Load_0pt2ml_P"~"HIV VL"),levels=c("CD4+ count","HIV VL")),
    timepoint=str_to_title(timepoint)
  )

for(i in 1:nrow(df)){
  df$n[i]<-sum(tmp$marker==df$marker[i] & tmp$group==df$group[i] & tmp$timepoint==df$timepoint[i] & !is.na(tmp$measurement))
  if(df$marker[i]=="CD4+ count"){df$k[i]<-sum(tmp$marker==df$marker[i] & tmp$group==df$group[i] & tmp$timepoint==df$timepoint[i] & !is.na(tmp$measurement) & tmp$measurement<200)}
  if(df$marker[i]=="HIV VL"){df$k[i]<-sum(tmp$marker==df$marker[i] & tmp$group==df$group[i] & tmp$timepoint==df$timepoint[i] & !is.na(tmp$measurement) & tmp$measurement>1000)}
  df$prop[i]<-df$k[i]/df$n[i]
  binomTmp<-binom.test(x=df$k[i],n=df$n[i])$conf.int
  df$propLow[i]<-binomTmp[1]
  df$propUpp[i]<-binomTmp[2]
  
  if(i %in% seq(2,nrow(df),by=2)){
    df$pValue[i-1]<-fisher.test(matrix(nrow=2,byrow=TRUE,c(df[i-1,"n"]-df[i-1,"k"],df[i-1,"k"],df[i,"n"]-df[i,"k"],df[i,"k"])))$p.value
  }
}

g3a<-df %>%
  dplyr::filter(marker=="CD4+ count") %>%
  dplyr::mutate(
    timepointNum=case_when(group=="Sub-/clinical CVD events"~timepointNum-0.08,group=="No sub-/clinical CVD events"~timepointNum+0.08),
    marker=case_when(marker=="CD4+ count"~"CD4+ count < 200",marker=="HIV VL"~"HIV VL > 1000")
    ) %>%
  ggplot(mapping=aes(x=timepointNum,y=prop,ymin=propLow,ymax=propUpp,col=group,label=ifelse(pValue<0.05,"*","n.s."))) +
  geom_point(size=2) +
  geom_errorbar(width=0.2,alpha=0.65) +
  geom_line(lwd=0.85) +
  geom_text(mapping=aes(x=timepointNum,y=ifelse(!is.na(pValue) & pValue<0.05,0.975,1.00),size=ifelse(!is.na(pValue) & pValue<0.05,14,12),fontface=ifelse(!is.na(pValue) & pValue<0.05,"bold","plain")),col="black") +
  theme_minimal() +
  facet_wrap(~marker,nrow=1) +
  scale_color_manual(values=c("steelblue","orange"),name="") +
  scale_x_continuous(breaks=c(0,6,12),labels=c("Enrolment","6 months","12 months"),name="") +
  scale_size_continuous(limits = c(1,14)) +
  ylab("Proportion of PLWH with uncontrolled HIV") +
  guides(size=FALSE) +
  ylim(c(0,1)) +
  theme(axis.text.x = element_text(angle=45,hjust=1,vjust=1), panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank(),text=element_text(size=16),legend.position = "bottom",legend.direction="vertical")

g3b<-df %>%
  dplyr::filter(marker=="HIV VL") %>%
  dplyr::mutate(
    timepointNum=case_when(group=="Sub-/clinical CVD events"~timepointNum-0.08,group=="No sub-/clinical CVD events"~timepointNum+0.08),
    marker=case_when(marker=="CD4+ count"~"CD4+ count < 200",marker=="HIV VL"~"HIV VL > 1000")
    ) %>%
  ggplot(mapping=aes(x=timepointNum,y=prop,ymin=propLow,ymax=propUpp,col=group,label=ifelse(pValue<0.05,"*","n.s."))) +
  geom_point(size=2) +
  geom_errorbar(width=0.2,alpha=0.65) +
  geom_line(lwd=0.85) +
  geom_text(mapping=aes(x=timepointNum,y=ifelse(!is.na(pValue) & pValue<0.05,0.975,1.00),size=ifelse(!is.na(pValue) & pValue<0.05,14,12),fontface=ifelse(!is.na(pValue) & pValue<0.05,"bold","plain")),col="black") +
  theme_minimal() +
  facet_wrap(~marker,nrow=1) +
  scale_color_manual(values=c("steelblue","orange"),name="") +
  scale_x_continuous(breaks=c(0,6,12),labels=c("Enrolment","6 months","12 months"),name="") +
  scale_size_continuous(limits = c(1,14)) +
  ylab("") +
  guides(size=FALSE) +
  ylim(c(0,1)) +
  theme(axis.text.x = element_text(angle=45,hjust=1,vjust=1), panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank(),text=element_text(size=16),legend.position = "bottom",legend.direction="vertical")

# dfRR<-data.frame(
#   group=c("500+","350-499","200-349","<200","<150","150-999","1000+"),
#   RR=c(1,NA,NA,NA,1,NA,NA),
#   RRlogSE=NA,
#   RRlow=NA,
#   RRupp=NA,
#   marker=factor(c(rep("CD4+ count",4),rep("HIV VL",3)),levels=c("CD4+ count","HIV VL")),
#   sig=NA
# )

dfRR<-data.frame(
  group=c("Full immune restoration","Partial immune restoration","Minimal immune restoration","Complete virological suppression","Virological resurgence","Incomplete virological suppression"),
  RR=c(1,NA,NA,1,NA,NA),
  RRlogSE=NA,
  RRlow=NA,
  RRupp=NA,
  marker=factor(c(rep("CD4+ count",3),rep("HIV VL",3)),levels=c("CD4+ count","HIV VL")),
  sig=NA
)


clinEventsTmpPLWH<-clinEventsTmpPLWH %>%
  dplyr::mutate(
    cd4=rhiccaDatWide$lims.CD4_Absolute_enrolment[match(pid,rhiccaDatWide$pid)],
    vl=rhiccaDatWide$lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment[match(pid,rhiccaDatWide$pid)],
    cd4Restoration=rhiccaDatWide$cd4Restoration[match(pid,rhiccaDatWide$pid)],
    vlSuppression=rhiccaDatWide$vlSuppression[match(pid,rhiccaDatWide$pid)]
  ) %>%
  dplyr::mutate(
    cd4_grp=case_when(
      cd4>=500~"500+",
      cd4>=350 & cd4<500~"350-499",
      cd4>=200 & cd4<350~"200-349",
      cd4<200~"<200"
    ),
    vl_grp=case_when(
      vl<150~"<150",
      vl>=150 & vl<1000~"150-999",
      vl>=1000~"1000+"
    )
  )

rhiccaDatWide<-rhiccaDatWide %>%
    dplyr::mutate(
    cd4_grp_enrolment=case_when(
      lims.CD4_Absolute_enrolment>=500~"500+",
      lims.CD4_Absolute_enrolment>=350 & lims.CD4_Absolute_enrolment<500~"350-499",
      lims.CD4_Absolute_enrolment>=200 & lims.CD4_Absolute_enrolment<350~"200-349",
      lims.CD4_Absolute_enrolment<200~"<200"
    ),
    vl_grp_enrolment=case_when(
      lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment<150~"<150",
      lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment>=150 & lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment<1000~"150-999",
      lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment>=1000~"1000+"
    )
  )

for(i in 1:nrow(dfRR)){
  if(dfRR$marker[i]=="CD4+ count"){
    if(dfRR$group[i]!="Full immune restoration"){
      a0=sum(!is.na(clinEventsTmpPLWH$cd4Restoration) & clinEventsTmpPLWH$cd4Restoration=="Full immune restoration")
      a1=sum(!is.na(clinEventsTmpPLWH$cd4Restoration) & clinEventsTmpPLWH$cd4Restoration==dfRR$group[i])
      t0=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.ptype=="Patient" & rhiccaDatWide$cd4Restoration=="Full immune restoration"],na.rm=TRUE))
      t1=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.ptype=="Patient" & rhiccaDatWide$cd4Restoration==dfRR$group[i]],na.rm=TRUE))
      if((a0>0 | a1>0) & t0>0 & t1>0){
        dfRR$RR[i]<-(a1/t1)/(a0/t0)
        dfRR$RRlogSE[i]<-sqrt(1/a0+1/a1)
        if(a1>0){
          dfRR$RRlow[i]<-exp(log(dfRR$RR[i])-qnorm(0.975)*dfRR$RRlogSE[i])
          dfRR$RRupp[i]<-exp(log(dfRR$RR[i])+qnorm(0.975)*dfRR$RRlogSE[i])
        }else{
          tmp<-rateratio.midp(x=c(a0,a1),y=c(t0,t1))
          dfRR$RRlow[i]<-tmp$measure["Exposed2","lower"]
          dfRR$RRupp[i]<-tmp$measure["Exposed2","upper"]
        }
        dfRR$sig[i]<-ifelse(dfRR$RRlow[i]<=1 & 1<=dfRR$RRupp[i],"n.s.","*")
      }
    }
  }
  if(dfRR$marker[i]=="HIV VL"){
    if(dfRR$group[i]!="Complete virological suppression"){
      a0=sum(!is.na(clinEventsTmpPLWH$vlSuppression) & clinEventsTmpPLWH$vlSuppression=="Complete virological suppression",dfRR$marker[i])
      a1=sum(!is.na(clinEventsTmpPLWH$vlSuppression) & clinEventsTmpPLWH$vlSuppression==dfRR$group[i])
      t0=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.ptype=="Patient" & rhiccaDatWide$vlSuppression=="Complete virological suppression"],na.rm=TRUE))
      t1=as.numeric(sum(rhiccaDatWide$obsTime[rhiccaDatWide$crf.ptype=="Patient" & rhiccaDatWide$vlSuppression==dfRR$group[i]],na.rm=TRUE))
      if((a0>0 | a1>0) & t0>0 & t1>0){
        dfRR$RR[i]<-(a1/t1)/(a0/t0)
        dfRR$RRlogSE[i]<-sqrt(1/a0+1/a1)
        if(a1>0){
          dfRR$RRlow[i]<-exp(log(dfRR$RR[i])-qnorm(0.975)*dfRR$RRlogSE[i])
          dfRR$RRupp[i]<-exp(log(dfRR$RR[i])+qnorm(0.975)*dfRR$RRlogSE[i])
        }else{
          tmp<-rateratio.midp(x=c(a0,a1),y=c(t0,t1))
          dfRR$RRlow[i]<-tmp$measure["Exposed2","lower"]
          dfRR$RRupp[i]<-tmp$measure["Exposed2","upper"]
        }
        dfRR$sig[i]<-ifelse(dfRR$RRlow[i]<=1 & 1<=dfRR$RRupp[i],"n.s.","*")
      }
    }
  }
}

g4a<-dfRR %>%
  dplyr::filter(marker=="CD4+ count") %>%
  dplyr::mutate(
    group=factor(case_when(
      group=="Full immune restoration"~"Full immune restoration (ref.)",
      TRUE~group
    ),levels=c("Full immune restoration (ref.)","Partial immune restoration","Minimal immune restoration"))
  ) %>%
  ggplot(mapping=aes(x=group,y=RR,ymin=RRlow,ymax=RRupp,label=sig)) +
  geom_point(size=3) +
  geom_errorbar(width=0.2,alpha=0.65) +
  geom_hline(yintercept=1,lty=2,col="darkgrey") +
  geom_text(mapping=aes(y=ifelse(sig=="*",32,35),size=ifelse(sig=="*",14,12),fontface=ifelse(sig=="*","bold","plain")),col="black") +
  theme_minimal() +
  ylab("Rate ratio for sub-/clinical events") +
  xlab("") +
  scale_size_continuous(limits = c(1,14)) +
  scale_y_continuous(trans="log10") +
  facet_wrap(~marker) +
  theme(axis.text.x = element_text(angle=45,hjust=1,vjust=1), panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank(),text=element_text(size=16),legend.position = "bottom") +
  guides(size=FALSE) +
  coord_cartesian(ylim=c(0.045,50))
  #labs(title="CD4+ count")

g4b<-dfRR %>%
  dplyr::filter(marker=="HIV VL") %>%
  dplyr::mutate(
    group=factor(case_when(
      group=="Complete virological suppression"~"Complete virological\nsuppression (ref.)",
      TRUE~group
    ),levels=c("Complete virological\nsuppression (ref.)","Virological resurgence","Incomplete virological suppression"))
  ) %>%
  ggplot(mapping=aes(x=group,y=RR,ymin=RRlow,ymax=RRupp,label=sig)) +
  geom_point(size=3) +
  geom_errorbar(width=0.2,alpha=0.65) +
  geom_hline(yintercept=1,lty=2,col="darkgrey") +
  geom_text(mapping=aes(y=ifelse(sig=="*",32,35),size=ifelse(sig=="*",14,12),fontface=ifelse(sig=="*","bold","plain")),col="black") +
  theme_minimal() +
  ylab("") +
  xlab("") +
  scale_size_continuous(limits = c(1,14)) +
  scale_y_continuous(trans="log10") +
  facet_wrap(~marker) +
  theme(axis.text.x = element_text(angle=45,hjust=1,vjust=1), panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank(),text=element_text(size=16),legend.position = "bottom") +
  guides(size=FALSE) +
  coord_cartesian(ylim=c(0.045,50))
  #labs(title="HIV VL")

# g2<-grid.arrange(g2a,g2b,nrow=1)
# 
gPanelB<-(g3a | g3b) / (g4a | g4b) + plot_annotation(tag_levels=list(c("E","F","G","H")))

gTopB<-ggplot(mapping=aes(x=0.5,y=1,label="PLWH")) + geom_text(size=16,face="bold") + theme_void()

pdf(width=7.5,height=12,file=paste(sep="_",outPrefix,"Fig2_HIVcontrolGroups_CD4_VL_RRofEvents_PLWH.pdf"))
(g3a | g3b) / (g4a | g4b) + plot_annotation(tag_levels=list(c("E","F","G","H")))
dev.off()

## quartz_off_screen 
##                 2

g1a<-g1a + theme(text=element_text(size=20))
g1b<-g1b + theme(text=element_text(size=20))
g2a<-g2a + theme(text=element_text(size=20))
g2b<-g2b + theme(text=element_text(size=20))
g3a<-g3a + theme(text=element_text(size=20))
g3b<-g3b + theme(text=element_text(size=20))
g4a<-g4a + theme(text=element_text(size=20))
g4b<-g4b + theme(text=element_text(size=20))


pdf(width=18,height=18,file=paste(sep="_",outPrefix,"Fig2_RRofEvents_PLWH.pdf"))
#grid.arrange(gPanelA,gPanelB,nrow=1)
#(gTopA | gTopB) / (g1a | g1b | g3a | g3b) / (g2a | g2b | g4a | g4b) + plot_layout(heights=c(0.1,0.45,0.45)) +
# plot_annotation(tag_levels=list(c("","","A","B","E","F","C","D","G","H")))
(g1a | g1b | g3a | g3b) / (g2a | g2b | g4a | g4b) + plot_layout(heights=c(0.5,0.5)) +
plot_annotation(tag_levels=list(c("A","B","E","F","C","D","G","H")))
dev.off()

## quartz_off_screen 
##                 2

10.0.0.2 Viral control

clinEventsTmpPLWH <- clinEventsTmpPLWH %>%
  dplyr::mutate(
    vlBaseline=rhiccaDatWide$lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment[match(pid,rhiccaDatWide$pid)],
    vl6mts=rhiccaDatWide$lims.HIV_Abbott_Viral_Load_0pt2ml_P_6months[match(pid,rhiccaDatWide$pid)],
    vl12mts=rhiccaDatWide$lims.HIV_Abbott_Viral_Load_0pt2ml_P_12months[match(pid,rhiccaDatWide$pid)],
  )

clinEventsTmpPLWH %>%
  kable() %>%
  kable_styling(full_width = FALSE)

pid	ptype	artnaive	artRegimen	gender	age_group	cmvStatus	vzvStatus	cmvStatus_2levels	vzvStatus_2levels	eGFR_existingCKD	category	age_group_2lvls	cd4	vl	cd4Restoration	vlSuppression	cd4_grp	vl_grp	vlBaseline	vl6mts	vl12mts
RE11077	PLWH	ART naive	ART: No change to dolutegravir	Male	[45,55)					no	CVD	<55	174	2318			<200	1000+	2318
RE1111F	PLWH	prior ART	ART: No change to dolutegravir	Male	[55,65)					yes	CVD	55+	63	56740			<200	1000+	56740
RE12045	PLWH	ART naive	ART: No change to dolutegravir	Male	[55,65)					no	CVD	55+	107	637935			<200	1000+	637935
RE1351K	PLWH	ART naive	ART: No change to dolutegravir	Male	[35,45)					no	CVD	<55	116	264938			<200	1000+	264938
RE1394Z	PLWH	ART naive	ART: No change to dolutegravir	Female	[35,45)					no	CVD	<55	98	2141148			<200	1000+	2141148
RE1504I	PLWH	prior ART	ART: Change to dolutegravir	Male	[45,55)					no	CVD	<55	288	36875	Full immune restoration	Complete virological suppression	200-349	1000+	36875	1	1
RE1571X	PLWH	prior ART	ART: Change to dolutegravir	Female	[55,65)	persistently low titres	persistently low titres	not persistently high titres	not persistently high titres	no	CVD	55+	149	20161	Partial immune restoration		<200	1000+	20161	1
RE16306	PLWH	ART naive	ART: No change to dolutegravir	Male	[45,55)					no	CVD	<55	27	418846			<200	1000+	418846
RE10445	PLWH	ART naive	ART: Change to dolutegravir	Female	[55,65)	persistently low titres	persistently low titres	not persistently high titres	not persistently high titres	no	CIMT	55+	261	117923	Partial immune restoration	Complete virological suppression	200-349	1000+	117923	1	1
RE10453	PLWH	ART naive	ART: Change to dolutegravir	Male	[45,55)	persistently high titres	persistently high titres	persistently high titres	persistently high titres	no	CIMT	<55	50	66601	Minimal immune restoration	Complete virological suppression	<200	1000+	66601	150	1
RE1080Y	PLWH	prior ART	ART: No change to dolutegravir	Male	[55,65)	persistently low titres	persistently high titres	not persistently high titres	persistently high titres	no	CIMT	55+	280	426	Partial immune restoration	Complete virological suppression	200-349	150-999	426	1	1
RE1096I	PLWH	prior ART	ART: No change to dolutegravir	Female	[55,65)	progression to high titres	progression to low titres	not persistently high titres	not persistently high titres	no	CIMT	55+	84	24083	Minimal immune restoration	Complete virological suppression	<200	1000+	24083	1	150
RE1100L	PLWH	ART naive	ART: Change to dolutegravir	Male	[55,65)					no	CIMT	55+	75	27112	Minimal immune restoration	Complete virological suppression	<200	1000+	27112	1	1
RE1137W	PLWH	prior ART	ART: Change to dolutegravir	Male	[35,45)	persistently low titres	persistently high titres	not persistently high titres	persistently high titres	no	CIMT	<55	223	8446	Partial immune restoration	Complete virological suppression	200-349	1000+	8446	1	150
RE1486R	PLWH	ART naive	ART: No change to dolutegravir	Female	[65,Inf)	progression to high titres	persistently low titres	not persistently high titres	not persistently high titres	no	CIMT	55+	173	9114	Minimal immune restoration	Complete virological suppression	<200	1000+	9114	1	1
RE1860F	PLWH	prior ART	ART: Change to dolutegravir	Female	[55,65)	persistently low titres	persistently high titres	not persistently high titres	persistently high titres	no	CIMT	55+	630	10077	Full immune restoration	Complete virological suppression	500+	1000+	10077	1	1
RE19439	PLWH	prior ART	ART: No change to dolutegravir	Female	[45,55)	persistently low titres	persistently high titres	not persistently high titres	persistently high titres	no	CIMT	<55	99	22927	Minimal immune restoration	Complete virological suppression	<200	1000+	22927	1	1
RE1008D	PLWH	prior ART	ART: No change to dolutegravir	Female	[35,45)	persistently high titres	progression to low titres	persistently high titres	not persistently high titres	no	PWV	<55	140	3306	Partial immune restoration	Complete virological suppression	<200	1000+	3306	1	150
RE10453	PLWH	ART naive	ART: Change to dolutegravir	Male	[45,55)	persistently high titres	persistently high titres	persistently high titres	persistently high titres	no	PWV	<55	50	66601	Minimal immune restoration	Complete virological suppression	<200	1000+	66601	150	1
RE11245	PLWH	prior ART	ART: Change to dolutegravir	Male	[45,55)	persistently low titres	persistently low titres	not persistently high titres	not persistently high titres	no	PWV	<55	561	2778	Partial immune restoration	Complete virological suppression	500+	1000+	2778	1	150
RE12053	PLWH	ART naive	ART: No change to dolutegravir	Male	[35,45)					no	PWV	<55	330	75204	Full immune restoration	Complete virological suppression	200-349	1000+	75204	1	1
RE1251S	PLWH	ART naive	ART: Change to dolutegravir	Male	[55,65)					no	PWV	55+	306	225959	Partial immune restoration	Complete virological suppression	200-349	1000+	225959	1	1
RE13738	PLWH	ART naive	ART: Change to dolutegravir	Female	[45,55)	persistently low titres	persistently low titres	not persistently high titres	not persistently high titres	no	PWV	<55	188	442092	Partial immune restoration	Complete virological suppression	<200	1000+	442092	1	1
RE13834	PLWH	ART naive	ART: No change to dolutegravir	Male	[35,45)	persistently high titres	persistently high titres	persistently high titres	persistently high titres	no	PWV	<55	62	611993	Minimal immune restoration		<200	1000+	611993	150
RE1424I	PLWH	ART naive	ART: Change to dolutegravir	Male	[35,45)	persistently low titres	progression to high titres	not persistently high titres	not persistently high titres	no	PWV	<55	132	261297	Partial immune restoration	Complete virological suppression	<200	1000+	261297	1	1
RE1459X	PLWH	ART naive	ART: Change to dolutegravir	Male	[55,65)	persistently low titres	persistently low titres	not persistently high titres	not persistently high titres	no	PWV	55+		9178	Full immune restoration	Complete virological suppression		1000+	9178	1	1
RE1557T	PLWH	ART naive	ART: Change to dolutegravir	Female	[45,55)	persistently low titres	persistently low titres	not persistently high titres	not persistently high titres	no	PWV	<55	620	3257	Partial immune restoration		500+	1000+	3257	198
RE19551	PLWH	prior ART	ART: Change to dolutegravir	Male	[35,45)					no	PWV	<55			Partial immune restoration	Complete virological suppression				1	1

10.0.1 Hypothetical vaccine trial

Given the above results, we can consider how large a vaccine efficacy trial we would need to run to demonstrate an effect of a VZV vaccine on CVD outcome (clinical and sub-clinical CIMT events).

The design for such an hypothetical trial would be a standard 2 parellel arms design. There are considerable degrees of freedom around lengths of study and what population to recruit. Below we list our assumptions - subject to modification. Please not the distinction between study arm (VZV vaccine or control) and zoster infection group (pVZV or not).

Study design: 2-arm, randomised controlled trial.
Study duration: 30 months (for pragmatic reasons: 6 months between vaccination and first zoster measurement, then 24 months of follow-up for CVD outcome).
Intervention: varicella zoster vaccine (most likely the recombinant vaccine as evaluated in https://doi.org/10.1093/cid/ciac549).
Endpoint: count outcome of the number of CVD events (clinical or sub-clinical CIMT as defined in earlier sections).
Assumptions (based on data from RHICCA):
- Prevalence of pVZV in control arm: 17.9% (in RHICCA 74 out of 413 PLWH with recorded VZV status has pVZV).
- Target vaccine efficacy range: 10% to 60%.
- Event rates in pVZV group: 31.160 per 1,000 person-years.
- Event rates in pVZV group: 5,653 per 1,000 person-years.
- Loss-to-follow-up over 30 months: 25% (this was 26.7% in RHICCA over 24 months; but we expect slightly less LTFU in a trial setting).
Analysis will consist of a simple Poisson test comparing the rate of events between study arms.
We will aim for 80% power at a 5% significance level.

With the above assumptions regarding pVZV prevalence and event rates in the pVZV and non-pVZV groups, this corresponds to an event rate of 10.219 events per 1,000 person-years in the vaccine arm and between 7.479 (60% efficacy) and 9.762 (10% efficacy) events per 1,000 person-years in the control group.

The design would need to recruit PLWH, initiated on ART at recruitment, age range 35-55 years of age, healthy except for HIV status. Participants would be recruited, consented, then randomised to zoster vaccine or control with the vaccine administered at the recruitment visit or shortly thereafter. VZV would be measured first 6 months after vaccination and then again 30 months after vaccination. The primary analysis would consider an intention-to-treat population with the full 2.5 years since vaccination counting towards to primary outcome of CVD events. A secondary analysis would consider the same population but only the final 2 years of follow-up.

Based on these assumptions, we get the sample sizes listed in 10.3 and shown on 10.1.

library(PASSED)

ssDf<-data.frame(
  VE=seq(0.1,0.6,by=0.1),
  prevalence=0.179,
  LTFU=0.25,
  followUpTime="2 years",
  rateCtrl=NA,
  rateVacc=NA,
  nPerGroup=NA,
  nPerGroupLTFU=NA,
  nTotal=NA
) %>%
  dplyr::mutate(
    rateCtrl=prevalence*31.16/1000+(1-prevalence)*5.653/1000,
    rateVacc=(1-VE)*prevalence*31.16/1000+(1-(1-VE)*prevalence)*5.653/1000,
  )

for(i in 1:nrow(ssDf)){
  ssDf$nPerGroup[i]<-ceiling(power_Poisson(power=0.8,equal.sample=TRUE,sig.level=0.05,lambda1=ssDf$rateCtrl[i],lambda2=ssDf$rateVacc[i],t1=2,t2=2,RR0=1,alternative="two.sided")$N)
}

ssDf<-ssDf %>%
  dplyr::mutate(
    nPerGroupLTFU=ceiling(nPerGroup/(1-LTFU)),
    nTotal=2*ceiling(nPerGroup/(1-LTFU))
  )

ssDf %>%
  kable(caption="Sample size for a hypothetical varicella zoster vaccine trial.",row.names=F,col.names=c("Vaccine efficacy","Prevalence of pVZV","LTFU","Follow-up time","Event rate in control arm (per person-year)","Event rate in vaccine arm (per person-year)","Sample size (per group, no LTFU)","Sample size (per group, with LTFU)","Sample size (total trial)")) %>%
  kable_styling(full_width=FALSE)

Table 10.3: Sample size for a hypothetical varicella zoster vaccine trial.
Vaccine efficacy	Prevalence of pVZV	LTFU	Follow-up time	Event rate in control arm (per person-year)	Event rate in vaccine arm (per person-year)	Sample size (per group, no LTFU)	Sample size (per group, with LTFU)	Sample size (total trial)
0.1	0.179	0.25	2 years	0.0102188	0.0097622	373555	498074	996148
0.2	0.179	0.25	2 years	0.0102188	0.0093056	90580	120774	241548
0.3	0.179	0.25	2 years	0.0102188	0.0088490	39004	52006	104012
0.4	0.179	0.25	2 years	0.0102188	0.0083925	21231	28308	56616
0.5	0.179	0.25	2 years	0.0102188	0.0079359	13133	17511	35022
0.6	0.179	0.25	2 years	0.0102188	0.0074793	8802	11736	23472

ssDf %>%
  ggplot(mapping=aes(x=VE,y=nTotal)) +
  geom_point(col="orange",size=2) +
  geom_line(col="orange",lwd=1) +
  theme_minimal() +
  xlab("Vaccine efficacy") +
  ylab("Total sample size") +
  labs(caption="pVZV prevalence in control arm = 17.9%.\nCVD event rate in pVZV group = 31.160/1,000 PY, CVD event rate in non-pVZV group = 5.653/1000 PY.\nLTFU = 25%.\nFollow-up period = 24 months.") +
  scale_y_continuous(trans="log10")

Figure 10.1: Sample size for a hypothetical vaccine trial as a function of vaccine efficacy.

# modClinEvents<-with(rhiccaDatWide_impMice.clinEvents,glm(numClinEventsPerParticipant~crf.ptype*crf.gender+crf.age+physio.waisthip_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment,family="quasipoisson"))
# 
# modClinEventsNoOnset<-with(rhiccaDatWide_impMice.clinEvents,glm(numClinEventsPerParticipantNoOnset~crf.ptype*crf.gender+crf.age+physio.waisthip_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+pwv.pwv_enrolment+pwvChange+cimt.cimt_enrolment+cimtChange,family="quasipoisson"))
# 
# modClinEventsNoPwvCimtAbpi<-with(rhiccaDatWide_impMice.clinEvents,glm(numClinEventsPerParticipantNoPwvCimtAbpi~crf.ptype*crf.gender+crf.age+physio.waisthip_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+pwv.pwv_enrolment+pwvChange+cimt.cimt_enrolment+cimtChange+cmvStatus+vzvStatus,family="quasipoisson"))
# 
# modClinEventsConfirmed<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipant~crf.ptype*crf.gender+crf.age+physio.waisthip_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+pwv.pwv_enrolment+pwvChange+cimt.cimt_enrolment+cimtChange,family="quasipoisson"))
# 
# modClinEventsConfirmedWithDerivedWithoutPwvCimtAbpi<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantWithDerivedWithoutPwvCimtAbpi~crf.ptype*crf.gender+crf.age+physio.waisthip_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+pwv.pwv_enrolment+pwvChange+cimt.cimt_enrolment+cimtChange,family="quasipoisson"))
# 
# modClinEventsPwvCimtAbpi<-with(rhiccaDatWide_impMice.clinEvents,glm(numClinEventsPwvCimtAbpi~crf.ptype*crf.gender+crf.age+physio.waisthip_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmv.cmv_6months+vzv.vzv_6months,family="quasipoisson"))
# 
# modClinEventsPwvCimt<-with(rhiccaDatWide_impMice.clinEvents,glm(numClinEventsPwvCimt~crf.ptype*crf.gender+crf.age+physio.waisthip_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus+vzvStatus,family="quasipoisson"))
# 
# modClinEventsCVD<-with(rhiccaDatWide_impMice.clinEvents,glm(numClinEventsPerParticipantCVD~crf.ptype+crf.gender+crf.age+physio.waisthip_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+lum.CD163_enrolment+lum.ICAM1_enrolment+pwv.pwv_enrolment+pwvChange+cimt.cimt_enrolment+cimtChange,family="quasipoisson"))
# 
# modClinEventsConfirmedCVD<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVD~crf.ptype*crf.gender+crf.age+physio.waisthip_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus+vzvStatus,family="quasipoisson"))
# 
# modClinEventsConfirmedCVD.simple<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVD~crf.ptype+crf.gender+crf.age,family="quasipoisson"))
# 
# modClinEventsConfirmedCVD.univariable<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVD~crf.ptype,family="quasipoisson"))
# 
# modClinEventsConfirmedCVD.univariable.CmzVzv<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVD~crf.ptype+cmvStatus+vzvStatus,family="quasipoisson"))
# 
# modClinEventsConfirmedCVD.simple.CmzVzv<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVD~crf.ptype+crf.gender+crf.age+cmvStatus+vzvStatus,family="quasipoisson"))

modClinEventsConfirmedCVDwithCIMT.simple<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMT~crf.ptype+crf.gender+crf.age,offset=obsTime/(1000*365.25),family="quasipoisson"))

modClinEventsConfirmedCVDwithCIMT.univariable.CmzVzv<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMT~crf.ptype+cmvStatus_2levels+vzvStatus_2levels,offset=obsTime/(1000*365.25),family="quasipoisson"))

modClinEventsConfirmedCVDwithCIMT.simple.CmzVzv<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMT~crf.ptype+crf.gender+crf.age+cmvStatus_2levels+vzvStatus_2levels,offset=obsTime/(1000*365.25),family="quasipoisson"))

modClinEventsConfirmedCVDwithCIMT.full<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMT~crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels,offset=obsTime/(1000*365.25),family="quasipoisson"))

modClinEventsConfirmedCVDwithCIMT.simple<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMT~crf.ptype+crf.gender+crf.age,offset=obsTime/(1000*365.25),family="quasipoisson"))

# modClinEventsConfirmedCVDwithCIMTPWV.simple<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMTPWV~crf.ptype+crf.gender+crf.age,family="quasipoisson"))
# 
# modClinEventsConfirmedCVDwithCIMTPWV.univariable.CmzVzv<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMTPWV~crf.ptype+cmvStatus+vzvStatus,family="quasipoisson"))
# 
# modClinEventsConfirmedCVDwithCIMTPWV.simple.CmzVzv<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantCVDwithCIMTPWV~crf.ptype+crf.gender+crf.age+cmvStatus+vzvStatus,family="quasipoisson"))


modClinEventsSystemic<-with(rhiccaDatWide_impMice.clinEvents,glm(numClinEventsPerParticipantSystemic~crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lum.CD163_enrolment+lum.ICAM1_enrolment+pwv.pwv_enrolment+pwvChange+cimt.cimt_enrolment+cimtChange,family="quasipoisson"))

modClinEventsConfirmedSystemic<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantSystemic~crf.ptype*crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels,family="quasipoisson"))

tab1<-summary(pool(modClinEventsConfirmedCVDwithCIMT.simple)) %>%
  as.data.frame() %>%
  dplyr::select(!term) %>%
  mutate(rownum=1:nrow(summary(pool(modClinEventsConfirmedCVDwithCIMT.simple)))) %>%
  mutate(estimate=exp(estimate)) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum))

tab2<-summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable.CmzVzv)) %>%
  as.data.frame() %>%
  dplyr::select(!term) %>%
  mutate(rownum=1:nrow(summary(pool(modClinEventsConfirmedCVDwithCIMT.univariable.CmzVzv)))) %>%
  mutate(estimate=exp(estimate)) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum))

tab3<-summary(pool(modClinEventsConfirmedCVDwithCIMT.simple.CmzVzv)) %>%
  as.data.frame() %>%
  dplyr::select(!term) %>%
  mutate(rownum=1:nrow(summary(pool(modClinEventsConfirmedCVDwithCIMT.simple.CmzVzv)))) %>%
  mutate(estimate=exp(estimate)) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum))

tab4<-summary(pool(modClinEventsConfirmedCVDwithCIMT.full)) %>%
  as.data.frame() %>%
  dplyr::select(!term) %>%
  mutate(rownum=1:nrow(summary(pool(modClinEventsConfirmedCVDwithCIMT.full)))) %>%
  mutate(estimate=exp(estimate)) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum))

tabAllMods<-rbind(tab1,tab2,tab3,tab4)

rnames<-c(
  c("Intercept","Participant type PLWH","Sex male","Age (in years)"),
  c("Intercept ","Participant type PLWH ","CMV: persistently high titres","VZV: persistently high titres"),
  c("Intercept  ","Participant type PLWH  ","Sex male ","Age (in years) ","CMV: persistently high titres ","VZV: persistently high titres "),
  c("Intercept   ","Participant type PLWH   ","Sex male  ","Age (in years)  ","BMI","Diastolic blood pressure (enrolment)","Systolic blood pressure (enrolment)","Glucose (enrolment)","Cholesterol (enrolment)","CMV: persistently high titres  ","VZV: persistently high titres  ")
)

tabAllMods %>%
  kableRn(caption="Results for unadjusted, minimally adjusted and fully adjusted Poisson regression models for the number of confirmed CVD clinical events and sub-clinical CIMT derived events per 1,000 person-years. Models fitted to multiply imputed data. The estimate column below shows the reference risk for the intercept row and relative risks associated with each variable.",digits=4,row_names=rnames,escape=F)  %>%
  kable_styling(full_width=F) %>%
  kableExtra::pack_rows(start_row=1,end_row=4,group_label="Minimally adjusted model without CMV / VZV status.") %>%
  kableExtra::pack_rows(start_row=5,end_row=8,group_label="Unadjusted (except for HIV status) model interrogating CMV / VZV status.") %>%
  kableExtra::pack_rows(start_row=9,end_row=14,group_label="Minimally adjusted model with CMV / VZV status.") %>%
  kableExtra::pack_rows(start_row=15,end_row=25,group_label="Fully adjusted model.")

Table 10.4: Results for unadjusted, minimally adjusted and fully adjusted Poisson regression models for the number of confirmed CVD clinical events and sub-clinical CIMT derived events per 1,000 person-years. Models fitted to multiply imputed data. The estimate column below shows the reference risk for the intercept row and relative risks associated with each variable.
	estimate	std.error	statistic	df	p.value
Minimally adjusted model without CMV / VZV status.
Intercept	0.0000	1.0319	-9.9050	889.9168	0.0000
Participant type PLWH	2.8423	0.5464	1.9118	889.9168	0.0562
Sex male	0.9062	0.4059	-0.2427	889.9168	0.8083
Age (in years)	1.1199	0.0144	7.8882	889.9168	0.0000
Unadjusted (except for HIV status) model interrogating CMV / VZV status.
Intercept	0.0131	0.5887	-7.3682	860.7692	0.0000
Participant type PLWH	1.3692	0.6314	0.4976	881.4587	0.6189
CMV: persistently high titres	0.8369	0.6297	-0.2828	323.5349	0.7775
VZV: persistently high titres	3.1192	0.5175	2.1984	154.8026	0.0294
Minimally adjusted model with CMV / VZV status.
Intercept	0.0000	1.0749	-9.7710	872.2838	0.0000
Participant type PLWH	2.8196	0.5511	1.8810	851.2829	0.0603
Sex male	0.7848	0.4115	-0.5890	874.5199	0.5560
Age (in years)	1.1218	0.0151	7.6209	872.1795	0.0000
CMV: persistently high titres	0.9584	0.5507	-0.0771	207.2478	0.9386
VZV: persistently high titres	2.7524	0.4480	2.2598	130.3784	0.0255
Fully adjusted model.
Intercept	0.0000	2.5688	-5.0366	861.8102	0.0000
Participant type PLWH	3.6283	0.6620	1.9467	808.9141	0.0519
Sex male	0.8698	0.4987	-0.2797	862.9852	0.7798
Age (in years)	1.1383	0.0223	5.8177	853.4860	0.0000
BMI	1.0328	0.0584	0.5522	854.2737	0.5810
Diastolic blood pressure (enrolment)	1.0505	0.0345	1.4266	869.3522	0.1541
Systolic blood pressure (enrolment)	0.9850	0.0182	-0.8295	862.9430	0.4071
Glucose (enrolment)	0.8906	0.1885	-0.6146	852.8536	0.5390
Cholesterol (enrolment)	0.8031	0.2063	-1.0633	856.4854	0.2879
CMV: persistently high titres	0.7173	0.6310	-0.5266	211.0778	0.5990
VZV: persistently high titres	2.9716	0.4836	2.2521	168.7923	0.0256

10.1 All participants

# helper functions
funFitAndSum_modClinEventsConfirmedCVD<-function(varString,nDigits=2){
  mod<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(as.formula(paste(sep="","numClinEventsPerParticipantCVD~",varString)),offset=obsTime/(1000*365.25),family="quasipoisson"))
  
  sumObj<-summary(pool(mod)) %>%
    as.data.frame() %>%
    mutate(rownum=1:nrow(summary(pool(mod)))) %>%
    mutate(
      estimateLink=estimate,
      estimate=round(digits=nDigits,exp(estimate)),
      rrCI=NA
    ) %>%
    mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=T),cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=F)))
  
  for(j in 1:nrow(sumObj)){
    sumObj$rrCI[j]<-paste(sep="","(",paste(collapse=",",format(nsmall=nDigits,round(digits=nDigits,c(exp(sumObj$estimateLink[j]-qnorm(0.975)*sumObj$std.error[j]),exp(sumObj$estimateLink[j]+qnorm(0.975)*sumObj$std.error[j]))))),")")
  }
  
  sumObj<-sumObj %>% dplyr::select(!estimateLink)
  
  return(sumObj)         
}

funFitAndSum_modClinEventsConfirmedCVDwithCIMT<-function(varString,nDigits=2){
  mod<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(as.formula(paste(sep="","numClinEventsPerParticipantCVDwithCIMT~",varString)),offset=obsTime/(1000*365.25),family="quasipoisson"))
  
  sumObj<-summary(pool(mod)) %>%
    as.data.frame() %>%
    mutate(rownum=1:nrow(summary(pool(mod)))) %>%
    mutate(
      estimateLink=estimate,
      estimate=format(nsmall=nDigits,round(digits=nDigits,exp(estimate))),
      rrCI=NA
    ) %>%
    mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=T),cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=F))) %>%
    mutate(p.value=gsub(p.value,pattern="0.000",replacement="&lt;0.001"))
  
  for(j in 1:nrow(sumObj)){
    sumObj$rrCI[j]<-paste(sep="","(",paste(collapse=",",format(nsmall=nDigits,round(digits=nDigits,c(exp(sumObj$estimateLink[j]-qnorm(0.975)*sumObj$std.error[j]),exp(sumObj$estimateLink[j]+qnorm(0.975)*sumObj$std.error[j]))))),")")
  }
  
  sumObj<-sumObj %>% dplyr::select(!estimateLink)
  
  return(sumObj)         
}

funFitAndSum_modClinEventsCIMTonly<-function(varString,nDigits=2){
  mod<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(as.formula(paste(sep="","numClinEventsPerParticipantCIMTonly~",varString)),offset=obsTime/(1000*365.25),family="quasipoisson"))
  
  sumObj<-summary(pool(mod)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(mod)))) %>%
  mutate(
    estimateLink=estimate,
    estimate=format(nsmall=nDigits,round(digits=nDigits,exp(estimate))),
    rrCI=NA
  ) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=T),cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=F))) %>%
    mutate(p.value=gsub(p.value,pattern="0.000",replacement="&lt;0.001"))
  
  for(j in 1:nrow(sumObj)){
    sumObj$rrCI[j]<-paste(sep="","(",paste(collapse=",",format(nsmall=nDigits,round(digits=nDigits,c(exp(sumObj$estimateLink[j]-qnorm(0.975)*sumObj$std.error[j]),exp(sumObj$estimateLink[j]+qnorm(0.975)*sumObj$std.error[j]))))),")")
  }
     
  sumObj<-sumObj %>% dplyr::select(!estimateLink)
  
  return(sumObj)         
}

funFitAndSum_modClinEventsConfirmedCVD.all<-funFitAndSum_modClinEventsConfirmedCVD
funFitAndSum_modClinEventsCIMTonly.all<-funFitAndSum_modClinEventsCIMTonly
funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT

# asymtomatic CIMT
tab.cimt<-tabAll.cimt %>%
  dplyr::select(!contains("hivadj")) %>%
  dplyr::filter(!grepl(term,pattern="CD4|HIV"))

# clinical CVD events
tab1.ptype<-funFitAndSum_modClinEventsConfirmedCVD("crf.ptype")
tab1.sex<-funFitAndSum_modClinEventsConfirmedCVD("crf.gender")
tab1.age<-funFitAndSum_modClinEventsConfirmedCVD("crf.age")
tab1.smoke<-funFitAndSum_modClinEventsConfirmedCVD("crf.smoke") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.alcohol<-funFitAndSum_modClinEventsConfirmedCVD("crf.alcohol") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.waisthip<-funFitAndSum_modClinEventsConfirmedCVD("physio.waisthip_enrolment")
tab1.bmi<-funFitAndSum_modClinEventsConfirmedCVD("physio.bmi_enrolment")
tab1.diabp<-funFitAndSum_modClinEventsConfirmedCVD("physio.bp_dia_enrolment")
tab1.sysbp<-funFitAndSum_modClinEventsConfirmedCVD("physio.bp_sys_enrolment")
tab1.glucose<-funFitAndSum_modClinEventsConfirmedCVD("lims.Glucose_enrolment")
tab1.cholesterol<-funFitAndSum_modClinEventsConfirmedCVD("lims.Cholesterol_enrolment")
tab1.ckd<-funFitAndSum_modClinEventsConfirmedCVD("eGFR_existingCKD") # not retained: no events in risk group --> ridiculous / unstable coefficient
#tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVD("cmvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
#tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVD("vzvStatus") # not retained as no events in some VZV groups --> ridiculous / unstable coefficient for that group
tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVD("cmvStatus_2levels")
tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVD("vzvStatus_2levels")

tab1<-rbind(tab1.ptype,
            tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd
            tab1.cmv,
            tab1.vzv)
tab1$term<-as.character(tab1$term)

tab2<-funFitAndSum_modClinEventsConfirmedCVD("crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels")
tab2$term<-as.character(tab2$term)

allRows<-setdiff(unique(c(tab1$term,tab2$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$rrCI.unadj<-tab1$rrCI[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.clinevents<-tabAll


# sub-clinical CIMT events only
tab1.ptype<-funFitAndSum_modClinEventsCIMTonly("crf.ptype")
tab1.sex<-funFitAndSum_modClinEventsCIMTonly("crf.gender")
tab1.age<-funFitAndSum_modClinEventsCIMTonly("crf.age")
tab1.smoke<-funFitAndSum_modClinEventsCIMTonly("crf.smoke") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.alcohol<-funFitAndSum_modClinEventsCIMTonly("crf.alcohol") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.waisthip<-funFitAndSum_modClinEventsCIMTonly("physio.waisthip_enrolment")
tab1.bmi<-funFitAndSum_modClinEventsCIMTonly("physio.bmi_enrolment")
tab1.diabp<-funFitAndSum_modClinEventsCIMTonly("physio.bp_dia_enrolment")
tab1.sysbp<-funFitAndSum_modClinEventsCIMTonly("physio.bp_sys_enrolment")
tab1.glucose<-funFitAndSum_modClinEventsCIMTonly("lims.Glucose_enrolment")
tab1.cholesterol<-funFitAndSum_modClinEventsCIMTonly("lims.Cholesterol_enrolment")
tab1.ckd<-funFitAndSum_modClinEventsCIMTonly("eGFR_existingCKD") # not retained: no events in risk group --> ridiculous / unstable coefficient
#tab1.cmv<-funFitAndSum_modClinEventsCIMTonly("cmvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
#tab1.vzv<-funFitAndSum_modClinEventsCIMTonly("vzvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
tab1.cmv<-funFitAndSum_modClinEventsCIMTonly("cmvStatus_2levels")
tab1.vzv<-funFitAndSum_modClinEventsCIMTonly("vzvStatus_2levels")

tab1<-rbind(tab1.ptype,
            tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv)
tab1$term<-as.character(tab1$term)

tab2<-funFitAndSum_modClinEventsCIMTonly("crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels")
tab2$term<-as.character(tab2$term)

allRows<-setdiff(unique(c(tab1$term,tab2$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$rrCI.unadj<-tab1$rrCI[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.subclineventscimt<-tabAll

# clinical CVD events + sub-clinical CIMT events
tab1.ptype<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.ptype")
tab1.sex<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.gender")
tab1.age<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.age")
tab1.smoke<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.smoke") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.alcohol<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.alcohol") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.waisthip<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("physio.waisthip_enrolment")
tab1.bmi<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("physio.bmi_enrolment")
tab1.diabp<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("physio.bp_dia_enrolment")
tab1.sysbp<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("physio.bp_sys_enrolment")
tab1.glucose<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("lims.Glucose_enrolment")
tab1.cholesterol<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("lims.Cholesterol_enrolment")
tab1.ckd<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("eGFR_existingCKD") # not retained: no events in risk group --> ridiculous / unstable coefficient
#tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("cmvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
#tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("vzvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("cmvStatus_2levels")
tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("vzvStatus_2levels")

tab1<-rbind(tab1.ptype,
            tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv)
tab1$term<-as.character(tab1$term)

tab2<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels")
tab2$term<-as.character(tab2$term)

allRows<-setdiff(unique(c(tab1$term,tab2$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$rrCI.unadj<-tab1$rrCI[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.clineventswithcimt<-tabAll

# combining all in a single table
#colnames(tab.cimt)<-paste(sep="_","cimt",colnames(tab.cimt))
colnames(tab.clinevents)<-paste(sep="_","clinevents",colnames(tab.clinevents))
colnames(tab.subclineventscimt)<-paste(sep="_","subclineventscimt",colnames(tab.subclineventscimt))
colnames(tab.clineventswithcimt)<-paste(sep="_","clineventwithcimt",colnames(tab.clineventswithcimt))

allRows<-unique(c(tab.clinevents$clinevents_term,tab.subclineventscimt$subclineventscimt_term,tab.clineventswithcimt$clineventswithcimt_term))
                
tabAll<-cbind(
  allRows,
  #tab.cimt[match(allRows,tab.cimt$cimt_term),] %>% dplyr::select(!contains("term")),
  tab.clinevents[match(allRows,tab.clinevents$clinevents_term),] %>% dplyr::select(!contains("term")),
  tab.subclineventscimt[match(allRows,tab.subclineventscimt$subclineventscimt_term),] %>% dplyr::select(!contains("term")),
  tab.clineventswithcimt[match(allRows,tab.clineventswithcimt$clineventwithcimt_term),] %>% dplyr::select(!contains("term"))
)

options(knitr.kable.NA = '')

rn<-c("Participant type: PLWH","Sex: male","Age (in years)","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose","Cholesterol","CMV: persistently high titres","VZV: persistently high titres")

tabAll %>%
  dplyr::select(!allRows) %>%
  kableRn(digits=3,escape=F,row_names=rn,col.names=c(rep(c("RR","95% CI","p"),2),rep(c("RR","95% CI","p"),4)),caption="Model summaries for models 1) fitted to clinical CVD events, 2) fitted to clinical CVD events only and 3) fitted to clinical CVD events and sub-clinical CIMT events (both unadjusted, single variable models and multiple regression models for all participants). The event count models are Poisson models and hence report rate ratios. Models fitted to multiply-imputed data from all participants.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  add_header_above(header=c(" "=1,"Unadjusted models" = 3, "Multiple regression model" = 3,"Unadjusted models" = 3, "Multiple regression model" = 3,"Unadjusted models" = 3, "Multiple regression model" = 3)) %>%
  add_header_above(header=c(" "=1,"Clinical CVD events" = 6, "Sub-clinical CIMT events" = 6,"Clinical CVD and sub-clinical CIMT events" = 6)) %>%
  kable_styling(full_width=F)

Table 10.5: Model summaries for models 1) fitted to clinical CVD events, 2) fitted to clinical CVD events only and 3) fitted to clinical CVD events and sub-clinical CIMT events (both unadjusted, single variable models and multiple regression models for all participants). The event count models are Poisson models and hence report rate ratios. Models fitted to multiply-imputed data from all participants.
	Clinical CVD events						Sub-clinical CIMT events						Clinical CVD and sub-clinical CIMT events
	Unadjusted models			Multiple regression model			Unadjusted models			Multiple regression model			Unadjusted models			Multiple regression model
	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p
Participant type: PLWH	2.10	( 0.26,16.64)	0.48	3.98	( 0.34,46.83)	0.27	1.18	(0.26,5.41)	0.83	3.73	( 1.37,10.16)	0.01	1.49	(0.44,5.01)	0.52	3.63	( 0.99,13.28)	0.05
Sex: male	2.31	( 0.48,11.05)	0.30	2.40	( 0.35,16.61)	0.38	0.79	(0.24,2.58)	0.70	0.33	(0.15,0.76)	0.01	1.22	(0.49,3.04)	0.66	0.87	(0.33,2.31)	0.78
Age (in years)	1.09	(1.03,1.14)	0.00	1.10	(1.02,1.19)	0.01	1.13	(1.09,1.16)	0.00	1.18	(1.14,1.23)	0.00	1.11	(1.08,1.14)	0.00	1.14	(1.09,1.19)	0.00
BMI	1.01	(0.89,1.15)	0.86	1.10	(0.94,1.28)	0.24	0.96	(0.84,1.10)	0.60	0.96	(0.86,1.07)	0.43	0.99	(0.90,1.08)	0.78	1.03	(0.92,1.16)	0.58
Diastolic blood pressure	1.03	(0.98,1.08)	0.20	1.02	(0.91,1.15)	0.74	1.04	(1.00,1.08)	0.07	1.08	(1.02,1.14)	0.01	1.03	(1.00,1.06)	0.03	1.05	(0.98,1.12)	0.15
Systolic blood pressure	1.02	(1.00,1.05)	0.06	1.00	(0.94,1.07)	0.93	1.03	(1.01,1.05)	0.01	0.97	(0.95,1.00)	0.05	1.03	(1.01,1.04)	0.00	0.99	(0.95,1.02)	0.41
Glucose	0.86	(0.43,1.71)	0.66	0.75	(0.32,1.72)	0.50	1.02	(0.78,1.33)	0.91	0.98	(0.80,1.21)	0.85	0.97	(0.74,1.29)	0.86	0.89	(0.62,1.29)	0.54
Cholesterol	0.81	(0.44,1.49)	0.50	0.65	(0.31,1.36)	0.25	1.25	(0.78,2.02)	0.35	0.96	(0.71,1.29)	0.77	1.05	(0.72,1.53)	0.82	0.80	(0.54,1.20)	0.29
CMV: persistently high titres	0.91	(0.12,6.72)	0.92	0.57	(0.06,5.62)	0.63	0.98	(0.21,4.52)	0.98	0.87	(0.34,2.23)	0.77	0.96	(0.29,3.21)	0.95	0.72	(0.21,2.47)	0.60
VZV: persistently high titres	1.37	(0.20,9.36)	0.75	1.22	(0.15,9.70)	0.85	5.46	( 1.57,18.91)	0.01	6.38	( 3.08,13.23)	0.00	3.16	(1.16,8.61)	0.03	2.97	(1.15,7.67)	0.03

colnames(tabAll)[colnames(tabAll)=="allRows"]<-"predictor"

tabAll.all<-tabAll[,c(1,11:13,17:19)]

#tabAll$cimt_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt_pvalue.unadj)))
#tabAll$cimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt_pvalue.adj)))
tabAll$clinevents_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$clinevents_pvalue.unadj)))
tabAll$clinevents_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$clinevents_pvalue.adj)))
tabAll$subclineventscimt_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$subclineventscimt_pvalue.unadj)))
tabAll$subclineventscimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$subclineventscimt_pvalue.adj)))
tabAll$clineventwithcimt_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$clineventwithcimt_pvalue.unadj)))
tabAll$clineventwithcimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$clineventwithcimt_pvalue.adj)))
write.csv(tabAll,row.names=F,file=paste(sep="_",outPrefix,"CIMT_CVDevents_AllParticipants_Table5.csv"))

10.2 PLWH

ttt<-mice::complete(rhiccaDatWide_impMice.clinEventsConfirmed,action = "long",include = TRUE) %>%
  dplyr::mutate(numClinEventsPerParticipantCVDwithPWV=numClinEventsPerParticipantPWVonly+numClinEventsPerParticipantCVD)

rhiccaDatWide_impMice.clinEventsConfirmed<-mice::as.mids(ttt)


# helper functions
funFitAndSum_modCimt<-function(varString,nDigits=2){
  mod<-with(rhiccaDatWide_impMice %>% dplyr::filter(crf.ptype=="Patient"),glm(as.formula(paste(sep="","cimtChange~cimt.cimt_enrolment",varString))))
  
  sumObj<-summary(pool(mod)) %>%
    as.data.frame() %>%
    mutate(rownum=1:nrow(summary(pool(mod)))) %>%
    mutate(
      estimateLink=estimate,
      estimate=format(nsmall=nDigits,round(digits=nDigits,exp(estimate))),
      rrCI=NA
    ) %>%
    mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=T),cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=F))) %>%
    mutate(p.value=gsub(p.value,pattern="0.00",replacement="&lt;0.01"))
  
  for(j in 1:nrow(sumObj)){
    sumObj$rrCI[j]<-paste(sep="","(",paste(collapse=",",format(nsmall=nDigits,round(digits=nDigits,c(exp(sumObj$estimateLink[j]-qnorm(0.975)*sumObj$std.error[j]),exp(sumObj$estimateLink[j]+qnorm(0.975)*sumObj$std.error[j]))))),")")
  }
  
  sumObj<-sumObj %>% dplyr::select(!estimateLink)
  
  return(sumObj)         
}

funFitAndSum_modClinEventsConfirmedCVD<-function(varString,nDigits=2){
  mod<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(as.formula(paste(sep="","numClinEventsPerParticipantCVD~",varString)),offset=obsTime/(1000*365.25),family="quasipoisson"))
  
  sumObj<-summary(pool(mod)) %>%
    as.data.frame() %>%
    mutate(rownum=1:nrow(summary(pool(mod)))) %>%
    mutate(
      estimateLink=estimate,
      estimate=format(nsmall=nDigits,round(digits=nDigits,exp(estimate))),
      rrCI=NA
    ) %>%
    mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=T),cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=F))) %>%
    mutate(p.value=gsub(p.value,pattern="0.000",replacement="&lt;0.001"))
  
  for(j in 1:nrow(sumObj)){
    sumObj$rrCI[j]<-paste(sep="","(",paste(collapse=",",format(nsmall=nDigits,round(digits=nDigits,c(exp(sumObj$estimateLink[j]-qnorm(0.975)*sumObj$std.error[j]),exp(sumObj$estimateLink[j]+qnorm(0.975)*sumObj$std.error[j]))))),")")
  }
  
  sumObj<-sumObj %>% dplyr::select(!estimateLink)
  
  return(sumObj)         
}

funFitAndSum_modClinEventsConfirmedCVDwithCIMT<-function(varString,nDigits=2){
  mod<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(as.formula(paste(sep="","numClinEventsPerParticipantCVDwithCIMT~",varString)),offset=obsTime/(1000*365.25),family="quasipoisson"))
  
  sumObj<-summary(pool(mod)) %>%
    as.data.frame() %>%
    mutate(rownum=1:nrow(summary(pool(mod)))) %>%
    mutate(
      estimateLink=estimate,
      estimate=format(nsmall=nDigits,round(digits=nDigits,exp(estimate))),
      rrCI=NA
    ) %>%
    mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=T),cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=F))) %>%
    mutate(p.value=gsub(p.value,pattern="0.000",replacement="&lt;0.001"))
  
  for(j in 1:nrow(sumObj)){
    sumObj$rrCI[j]<-paste(sep="","(",paste(collapse=",",format(nsmall=nDigits,round(digits=nDigits,c(exp(sumObj$estimateLink[j]-qnorm(0.975)*sumObj$std.error[j]),exp(sumObj$estimateLink[j]+qnorm(0.975)*sumObj$std.error[j]))))),")")
  }
  
  sumObj<-sumObj %>% dplyr::select(!estimateLink)
  
  return(sumObj)         
}

funFitAndSum_modClinEventsConfirmedCVDwithPWV<-function(varString,filterPLWH=TRUE,nDigits=2){
  if(filterPLWH){
    mod<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(as.formula(paste(sep="","numClinEventsPerParticipantCVDwithPWV~",varString)),offset=obsTime/(1000*365.25),family="quasipoisson"))
  }else{
    mod<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(as.formula(paste(sep="","numClinEventsPerParticipantCVDwithPWV~",varString)),offset=obsTime/(1000*365.25),family="quasipoisson"))
  }
  
  sumObj<-summary(pool(mod)) %>%
    as.data.frame() %>%
    mutate(rownum=1:nrow(summary(pool(mod)))) %>%
    mutate(
      estimateLink=estimate,
      estimate=format(nsmall=nDigits,round(digits=nDigits,exp(estimate))),
      rrCI=NA
    ) %>%
    mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=T),cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=F))) %>%
    mutate(p.value=gsub(p.value,pattern="0.000",replacement="&lt;0.001"))
  
  for(j in 1:nrow(sumObj)){
    sumObj$rrCI[j]<-paste(sep="","(",paste(collapse=",",format(nsmall=nDigits,round(digits=nDigits,c(exp(sumObj$estimateLink[j]-qnorm(0.975)*sumObj$std.error[j]),exp(sumObj$estimateLink[j]+qnorm(0.975)*sumObj$std.error[j]))))),")")
  }
  
  sumObj<-sumObj %>% dplyr::select(!estimateLink)
  
  return(sumObj)         
}

funFitAndSum_modClinEventsCIMTonly<-function(varString,nDigits=2){
  mod<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>%  dplyr::filter(crf.ptype=="Patient"),glm(as.formula(paste(sep="","numClinEventsPerParticipantCIMTonly~",varString)),offset=obsTime/(1000*365.25),family="quasipoisson"))
  
  sumObj<-summary(pool(mod)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(mod)))) %>%
  mutate(
    estimateLink=estimate,
    estimate=format(nsmall=nDigits,round(digits=nDigits,exp(estimate))),
    rrCI=NA
  ) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=T),cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=F))) %>%
    mutate(p.value=gsub(p.value,pattern="0.000",replacement="&lt;0.001"))
  
  for(j in 1:nrow(sumObj)){
    sumObj$rrCI[j]<-paste(sep="","(",paste(collapse=",",format(nsmall=nDigits,round(digits=nDigits,c(exp(sumObj$estimateLink[j]-qnorm(0.975)*sumObj$std.error[j]),exp(sumObj$estimateLink[j]+qnorm(0.975)*sumObj$std.error[j]))))),")")
  }
     
  sumObj<-sumObj %>% dplyr::select(!estimateLink)
  
  return(sumObj)         
}

funFitAndSum_modClinEventsPWVonly<-function(varString,filterPLWH=TRUE,nDigits=2){
  if(filterPLWH){
    mod<-with(rhiccaDatWide_impMice.clinEventsConfirmed %>%  dplyr::filter(crf.ptype=="Patient"),glm(as.formula(paste(sep="","numClinEventsPerParticipantPWVonly~",varString)),offset=obsTime/(1000*365.25),family="quasipoisson"))
  }else{
    mod<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(as.formula(paste(sep="","numClinEventsPerParticipantPWVonly~",varString)),offset=obsTime/(1000*365.25),family="quasipoisson"))
  }
  
  sumObj<-summary(pool(mod)) %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(pool(mod)))) %>%
  mutate(
    estimateLink=estimate,
    estimate=format(nsmall=nDigits,round(digits=nDigits,exp(estimate))),
    rrCI=NA
  ) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=T),cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=F))) %>%
    mutate(p.value=gsub(p.value,pattern="0.000",replacement="&lt;0.001"))
  
  for(j in 1:nrow(sumObj)){
    sumObj$rrCI[j]<-paste(sep="","(",paste(collapse=",",format(nsmall=nDigits,round(digits=nDigits,c(exp(sumObj$estimateLink[j]-qnorm(0.975)*sumObj$std.error[j]),exp(sumObj$estimateLink[j]+qnorm(0.975)*sumObj$std.error[j]))))),")")
  }
     
  sumObj<-sumObj %>% dplyr::select(!estimateLink)
  
  return(sumObj)         
}

funFitAndSum_modClinEventsConfirmedCVD.hiv<-funFitAndSum_modClinEventsConfirmedCVD
funFitAndSum_modClinEventsCIMTonly.hiv<-funFitAndSum_modClinEventsCIMTonly
funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT


# asymptomatic CIMT
tab1.baseline<-funFitAndSum_modCimt("")
tab1.sex<-funFitAndSum_modCimt("+crf.gender")
tab1.age<-funFitAndSum_modCimt("+crf.age")
tab1.smoke<-funFitAndSum_modCimt("+crf.smoke") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.alcohol<-funFitAndSum_modCimt("+crf.alcohol") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.waisthip<-funFitAndSum_modCimt("+physio.waisthip_enrolment")
tab1.bmi<-funFitAndSum_modCimt("+physio.bmi_enrolment")
tab1.diabp<-funFitAndSum_modCimt("+physio.bp_dia_enrolment")
tab1.sysbp<-funFitAndSum_modCimt("+physio.bp_sys_enrolment")
tab1.glucose<-funFitAndSum_modCimt("+lims.Glucose_enrolment")
tab1.cholesterol<-funFitAndSum_modCimt("+lims.Cholesterol_enrolment")
tab1.ckd<-funFitAndSum_modCimt("+eGFR_existingCKD") # not retained: no events in risk group --> ridiculous / unstable coefficient
#tab1.cmv<-funFitAndSum_modCimt("+cmvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
#tab1.vzv<-funFitAndSum_modCimt("+vzvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
tab1.cmv<-funFitAndSum_modCimt("+cmvStatus_2levels") 
tab1.vzv<-funFitAndSum_modCimt("+vzvStatus_2levels")
tab1.cd4base<-funFitAndSum_modCimt("+lims.CD4_Absolute_enrolment")
tab1.cd4change<-funFitAndSum_modCimt("+lims.CD4_Absolute_changeFromBaselineAt24Months")
tab1.vl<-funFitAndSum_modCimt("+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")

tab1<-rbind(tab1.baseline,
            tab1.sex %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.age %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.smoke %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.alcohol %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.bmi %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.diabp %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.sysbp %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.glucose %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.cholesterol %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.ckd %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.cmv %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.vzv %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.cd4base %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.cd4change %>% dplyr::filter(term!="cimt.cimt_enrolment"),
            tab1.vl %>% dplyr::filter(term!="cimt.cimt_enrolment")
)
tab1$term<-as.character(tab1$term)

tab2<-tabAll.cimt %>%
  dplyr::select(c(term,contains("hivadj"))) %>%
  dplyr::filter(!grepl(term,pattern="ptype"))

allRows<-setdiff(unique(c(tab1$term,tab2$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,coef.unadj=NA,SE.unadj=NA,pvalue.unadj=NA,coef.adj=NA,SE.adj=NA,pvalue.adj=NA)

tabAll$coef.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$SE.unadj<-tab1$std.error[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$coef.adj<-tab2$coefficient.hivadj[match(tabAll$term,tab2$term)]
tabAll$SE.adj<-tab2$SE.hivadj[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$pvalue.hivadj[match(tabAll$term,tab2$term)]

tab.cimt<-tabAll

# clinical CVD events
tab1.sex<-funFitAndSum_modClinEventsConfirmedCVD("crf.gender")
tab1.age<-funFitAndSum_modClinEventsConfirmedCVD("crf.age")
tab1.smoke<-funFitAndSum_modClinEventsConfirmedCVD("crf.smoke") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.alcohol<-funFitAndSum_modClinEventsConfirmedCVD("crf.alcohol") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.waisthip<-funFitAndSum_modClinEventsConfirmedCVD("physio.waisthip_enrolment")
tab1.bmi<-funFitAndSum_modClinEventsConfirmedCVD("physio.bmi_enrolment")
tab1.diabp<-funFitAndSum_modClinEventsConfirmedCVD("physio.bp_dia_enrolment")
tab1.sysbp<-funFitAndSum_modClinEventsConfirmedCVD("physio.bp_sys_enrolment")
tab1.glucose<-funFitAndSum_modClinEventsConfirmedCVD("lims.Glucose_enrolment")
tab1.cholesterol<-funFitAndSum_modClinEventsConfirmedCVD("lims.Cholesterol_enrolment")
tab1.ckd<-funFitAndSum_modClinEventsConfirmedCVD("eGFR_existingCKD") # not retained: no events in risk group --> ridiculous / unstable coefficient
#tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVD("cmvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
#tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVD("vzvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVD("cmvStatus_2levels")
tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVD("vzvStatus_2levels")
tab1.cd4base<-funFitAndSum_modClinEventsConfirmedCVD("lims.CD4_Absolute_enrolment")
tab1.cd4change<-funFitAndSum_modClinEventsConfirmedCVD("lims.CD4_Absolute_changeFromBaselineAt24Months")
tab1.vl<-funFitAndSum_modClinEventsConfirmedCVD("log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")

tab1<-rbind(tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv,
            tab1.cd4base,
            tab1.cd4change,
            tab1.vl)
tab1$term<-as.character(tab1$term)

tab1.nocd4<-rbind(tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv)
tab1.nocd4$term<-as.character(tab1.nocd4$term)

tab2<-funFitAndSum_modClinEventsConfirmedCVD("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")
tab2$term<-as.character(tab2$term)

tab2.nocd4<-funFitAndSum_modClinEventsConfirmedCVD("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels")
tab2.nocd4$term<-as.character(tab2.nocd4$term)

allRows<-setdiff(unique(c(tab1$term,tab2$term)),"(Intercept)")

allRows.nocd4<-setdiff(unique(c(tab1.nocd4$term,tab2.nocd4$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$rrCI.unadj<-tab1$rrCI[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.clinevents<-tabAll

tabAll.nocd4<-data.frame(term=allRows.nocd4,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll.nocd4$rr.unadj<-tab1.nocd4$estimate[match(tabAll.nocd4$term,tab1$term)]
tabAll.nocd4$rrCI.unadj<-tab1.nocd4$rrCI[match(tabAll.nocd4$term,tab1$term)]
tabAll.nocd4$pvalue.unadj<-tab1.nocd4$p.value[match(tabAll.nocd4$term,tab1$term)]

tabAll.nocd4$rr.adj<-tab2.nocd4$estimate[match(tabAll.nocd4$term,tab2.nocd4$term)]
tabAll.nocd4$rrCI.adj<-tab2.nocd4$rrCI[match(tabAll.nocd4$term,tab2.nocd4$term)]
tabAll.nocd4$pvalue.adj<-tab2.nocd4$p.value[match(tabAll.nocd4$term,tab2.nocd4$term)]

tab.clinevents.nocd4<-tabAll.nocd4

# sub-clinical CIMT events only
tab1.sex<-funFitAndSum_modClinEventsCIMTonly("crf.gender")
tab1.age<-funFitAndSum_modClinEventsCIMTonly("crf.age")
tab1.smoke<-funFitAndSum_modClinEventsCIMTonly("crf.smoke") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.alcohol<-funFitAndSum_modClinEventsCIMTonly("crf.alcohol") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.waisthip<-funFitAndSum_modClinEventsCIMTonly("physio.waisthip_enrolment")
tab1.bmi<-funFitAndSum_modClinEventsCIMTonly("physio.bmi_enrolment")
tab1.diabp<-funFitAndSum_modClinEventsCIMTonly("physio.bp_dia_enrolment")
tab1.sysbp<-funFitAndSum_modClinEventsCIMTonly("physio.bp_sys_enrolment")
tab1.glucose<-funFitAndSum_modClinEventsCIMTonly("lims.Glucose_enrolment")
tab1.cholesterol<-funFitAndSum_modClinEventsCIMTonly("lims.Cholesterol_enrolment")
tab1.ckd<-funFitAndSum_modClinEventsCIMTonly("eGFR_existingCKD") # not retained: no events in risk group --> ridiculous / unstable coefficient
#tab1.cmv<-funFitAndSum_modClinEventsCIMTonly("cmvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
#tab1.vzv<-funFitAndSum_modClinEventsCIMTonly("vzvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
tab1.cmv<-funFitAndSum_modClinEventsCIMTonly("cmvStatus_2levels")
tab1.vzv<-funFitAndSum_modClinEventsCIMTonly("vzvStatus_2levels")
tab1.cd4base<-funFitAndSum_modClinEventsCIMTonly("lims.CD4_Absolute_enrolment")
tab1.cd4change<-funFitAndSum_modClinEventsCIMTonly("lims.CD4_Absolute_changeFromBaselineAt24Months")
tab1.vl<-funFitAndSum_modClinEventsCIMTonly("log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")

tab1<-rbind(tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv,
            tab1.cd4base,
            tab1.cd4change,
            tab1.vl)
tab1$term<-as.character(tab1$term)

tab1.nocd4<-rbind(tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv)
tab1.nocd4$term<-as.character(tab1.nocd4$term)

tab2<-funFitAndSum_modClinEventsCIMTonly("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")
tab2$term<-as.character(tab2$term)

tab2.nocd4<-funFitAndSum_modClinEventsCIMTonly("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels")
tab2.nocd4$term<-as.character(tab2.nocd4$term)

allRows<-setdiff(unique(c(tab1$term,tab2$term)),"(Intercept)")

allRows.nocd4<-setdiff(unique(c(tab1.nocd4$term,tab2.nocd4$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$rrCI.unadj<-tab1$rrCI[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tabAll.nocd4<-data.frame(term=allRows.nocd4,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll.nocd4$rr.unadj<-tab1.nocd4$estimate[match(tabAll.nocd4$term,tab1.nocd4$term)]
tabAll.nocd4$rrCI.unadj<-tab1.nocd4$rrCI[match(tabAll.nocd4$term,tab1.nocd4$term)]
tabAll.nocd4$pvalue.unadj<-tab1.nocd4$p.value[match(tabAll.nocd4$term,tab1.nocd4$term)]

tabAll.nocd4$rr.adj<-tab2.nocd4$estimate[match(tabAll.nocd4$term,tab2.nocd4$term)]
tabAll.nocd4$rrCI.adj<-tab2.nocd4$rrCI[match(tabAll.nocd4$term,tab2.nocd4$term)]
tabAll.nocd4$pvalue.adj<-tab2.nocd4$p.value[match(tabAll.nocd4$term,tab2.nocd4$term)]

tab.subclineventscimt<-tabAll

tab.subclineventscimt.nocd4<-tabAll.nocd4

# clinical CVD events + sub-clinical CIMT events
tab1.sex<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.gender")
tab1.age<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.age")
tab1.smoke<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.smoke") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.alcohol<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.alcohol") # not retained: no events in risk group --> ridiculous / unstable coefficient
tab1.waisthip<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("physio.waisthip_enrolment")
tab1.bmi<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("physio.bmi_enrolment")
tab1.diabp<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("physio.bp_dia_enrolment")
tab1.sysbp<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("physio.bp_sys_enrolment")
tab1.glucose<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("lims.Glucose_enrolment")
tab1.cholesterol<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("lims.Cholesterol_enrolment")
tab1.ckd<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("eGFR_existingCKD") # not retained: no events in risk group --> ridiculous / unstable coefficient
#tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("cmvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
#tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("vzvStatus") # not retained as no events in some CMV groups --> ridiculous / unstable coefficient for that group
tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("cmvStatus_2levels")
tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("vzvStatus_2levels")
tab1.cd4base<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("lims.CD4_Absolute_enrolment")
tab1.cd4change<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("lims.CD4_Absolute_changeFromBaselineAt24Months")
tab1.vl<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")

tab1<-rbind(tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv,
            tab1.cd4base,
            tab1.cd4change,
            tab1.vl)
tab1$term<-as.character(tab1$term)

tab1.nocd4<-rbind(tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv)
tab1.nocd4$term<-as.character(tab1.nocd4$term)

tab2<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")
tab2$term<-as.character(tab2$term)

tab2.nocd4<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels")
tab2.nocd4$term<-as.character(tab2.nocd4$term)

allRows<-setdiff(unique(c(tab1$term,tab2$term)),"(Intercept)")

allRows.nocd4<-setdiff(unique(c(tab1.nocd4$term,tab2.nocd4$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$rrCI.unadj<-tab1$rrCI[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.clineventswithcimt<-tabAll

tabAll.nocd4<-data.frame(term=allRows.nocd4,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll.nocd4$rr.unadj<-tab1.nocd4$estimate[match(tabAll.nocd4$term,tab1.nocd4$term)]
tabAll.nocd4$rrCI.unadj<-tab1.nocd4$rrCI[match(tabAll.nocd4$term,tab1.nocd4$term)]
tabAll.nocd4$pvalue.unadj<-tab1.nocd4$p.value[match(tabAll.nocd4$term,tab1.nocd4$term)]

tabAll.nocd4$rr.adj<-tab2.nocd4$estimate[match(tabAll.nocd4$term,tab2.nocd4$term)]
tabAll.nocd4$rrCI.adj<-tab2.nocd4$rrCI[match(tabAll.nocd4$term,tab2.nocd4$term)]
tabAll.nocd4$pvalue.adj<-tab2.nocd4$p.value[match(tabAll.nocd4$term,tab2.nocd4$term)]

tab.clineventswithcimt.nocd4<-tabAll.nocd4

# combining all in a single table
#colnames(tab.cimt)<-paste(sep="_","cimt",colnames(tab.cimt))
colnames(tab.clinevents)<-paste(sep="_","clinevents",colnames(tab.clinevents))
colnames(tab.subclineventscimt)<-paste(sep="_","subclineventscimt",colnames(tab.subclineventscimt))
colnames(tab.clineventswithcimt)<-paste(sep="_","clineventwithcimt",colnames(tab.clineventswithcimt))

allRows<-unique(c(tab.clinevents$subclineventscimt_term,tab.subclineventscimt$subclineventscimt_term,tab.clineventswithcimt$clineventswithcimt_term))
                
tabAll<-cbind(
  allRows,
  #tab.cimt[match(allRows,tab.cimt$cimt_term),] %>% dplyr::select(!contains("term")),
  tab.clinevents[match(allRows,tab.clinevents$clinevents_term),] %>% dplyr::select(!contains("term")),
  tab.subclineventscimt[match(allRows,tab.subclineventscimt$subclineventscimt_term),] %>% dplyr::select(!contains("term")),
  tab.clineventswithcimt[match(allRows,tab.clineventswithcimt$clineventwithcimt_term),] %>% dplyr::select(!contains("term"))
)

colnames(tab.clinevents.nocd4)<-paste(sep="_","clinevents",colnames(tab.clinevents.nocd4))
colnames(tab.subclineventscimt.nocd4)<-paste(sep="_","subclineventscimt",colnames(tab.subclineventscimt.nocd4))
colnames(tab.clineventswithcimt.nocd4)<-paste(sep="_","clineventwithcimt",colnames(tab.clineventswithcimt.nocd4))

allRows.nocd4<-unique(c(tab.clinevents.nocd4$clinevents_term,tab.subclineventscimt.nocd4$subclineventscimt_term,tab.clineventswithcimt.nocd4$clineventswithcimt_term))
                
tabAll.nocd4<-cbind(
  allRows.nocd4,
  #tab.cimt[match(allRows,tab.cimt$cimt_term),] %>% dplyr::select(!contains("term")),
  tab.clinevents.nocd4[match(allRows.nocd4,tab.clinevents.nocd4$clinevents_term),] %>% dplyr::select(!contains("term")),
  tab.subclineventscimt.nocd4[match(allRows.nocd4,tab.subclineventscimt.nocd4$subclineventscimt_term),] %>% dplyr::select(!contains("term")),
  tab.clineventswithcimt.nocd4[match(allRows.nocd4,tab.clineventswithcimt.nocd4$clineventwithcimt_term),] %>% dplyr::select(!contains("term"))
)

options(knitr.kable.NA = '')

rn<-c("Sex: male","Age (in years)","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose","Cholesterol","CMV: persistently high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (change from baseline at 24 months)","log viral load (baseline)")

tabAll %>%
  dplyr::select(!allRows) %>%
  kableRn(digits=3,escape=F,row_names=rn,col.names=c(rep(c("RR","95% CI","p"),2),rep(c("RR","95% CI","p"),4)),caption="Model summaries for models for PLWH only 1) fitted to clinical CVD events only, 2) fitted to sub-clinical CIMT events only and 3) fitted to clinical CVD events and sub-clinical CIMT events (both unadjusted, single variable models and multiple regression models). The event count models are quasi-Poisson models and hence report rate ratios. Models fitted to multiply-imputed data from PLWH only.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  add_header_above(header=c(" "=1,"Unadjusted models" = 3, "Multiple regression model" = 3,"Unadjusted models" = 3, "Multiple regression model" = 3,"Unadjusted models" = 3, "Multiple regression model" = 3)) %>%
  add_header_above(header=c(" "=1,"Clinical CVD events" = 6, "Sub-clinical CIMT events" = 6,"Clinical CVD and sub-clinical CIMT events" = 6)) %>%
  kable_styling(full_width=F)

Table 10.6: Model summaries for models for PLWH only 1) fitted to clinical CVD events only, 2) fitted to sub-clinical CIMT events only and 3) fitted to clinical CVD events and sub-clinical CIMT events (both unadjusted, single variable models and multiple regression models). The event count models are quasi-Poisson models and hence report rate ratios. Models fitted to multiply-imputed data from PLWH only.
	Clinical CVD events						Sub-clinical CIMT events						Clinical CVD and sub-clinical CIMT events
	Unadjusted models			Multiple regression model			Unadjusted models			Multiple regression model			Unadjusted models			Multiple regression model
	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p
Sex: male	1.79	(0.36,8.80)	0.48	2.27	( 0.43,11.99)	0.33	0.48	(0.13,1.76)	0.27	0.16	(0.06,0.40)	0.00	0.85	(0.33,2.21)	0.74	0.68	(0.26,1.77)	0.42
Age (in years)	1.08	(1.02,1.15)	0.01	1.10	(1.03,1.18)	0.01	1.13	(1.09,1.17)	0.00	1.18	(1.13,1.24)	0.00	1.11	(1.07,1.14)	0.00	1.12	(1.07,1.17)	0.00
BMI	1.03	(0.90,1.18)	0.66	1.09	(0.95,1.25)	0.21	1.02	(0.89,1.17)	0.75	1.04	(0.94,1.15)	0.45	1.03	(0.93,1.13)	0.59	1.05	(0.95,1.16)	0.32
Diastolic blood pressure	1.01	(0.96,1.07)	0.61	1.04	(0.93,1.17)	0.47	1.04	(1.00,1.08)	0.04	1.06	(1.00,1.13)	0.05	1.03	(1.00,1.06)	0.07	1.04	(0.97,1.12)	0.26
Systolic blood pressure	1.01	(0.98,1.04)	0.38	0.99	(0.93,1.05)	0.73	1.03	(1.01,1.05)	0.01	0.99	(0.96,1.02)	0.69	1.02	(1.01,1.04)	0.01	0.99	(0.96,1.03)	0.72
Glucose	0.82	(0.37,1.78)	0.61	0.70	(0.35,1.41)	0.32	1.03	(0.80,1.33)	0.82	1.00	(0.83,1.21)	0.97	0.98	(0.74,1.31)	0.90	0.90	(0.62,1.29)	0.56
Cholesterol	0.83	(0.41,1.68)	0.61	0.74	(0.37,1.48)	0.39	1.23	(0.70,2.18)	0.47	0.65	(0.45,0.94)	0.02	1.04	(0.67,1.63)	0.86	0.74	(0.48,1.14)	0.17
CMV: persistently high titres	0.83	(0.11,6.34)	0.85	0.44	(0.06,3.20)	0.42	0.99	(0.21,4.75)	0.99	0.87	(0.33,2.32)	0.78	0.93	(0.27,3.19)	0.90	0.68	(0.21,2.22)	0.53
VZV: persistently high titres	0.03	(0.00, Inf)	1.00	0.03	(0.00, Inf)	1.00	7.95	( 1.87,33.89)	0.01	13.84	( 5.93,32.31)	0.00	2.96	(0.98,8.96)	0.06	3.08	(1.16,8.17)	0.03
CD4 (baseline)	0.99	(0.99,1.00)	0.06	0.99	(0.99,1.00)	0.08	1.00	(0.99,1.00)	0.66	1.00	(1.00,1.00)	0.31	1.00	(0.99,1.00)	0.14	1.00	(0.99,1.00)	0.11
CD4 (change from baseline at 24 months)	1.00	(1.00,1.01)	0.73	1.00	(0.99,1.01)	0.63	1.00	(1.00,1.00)	0.88	1.00	(1.00,1.00)	0.54	1.00	(1.00,1.00)	0.88	1.00	(1.00,1.00)	0.82
log viral load (baseline)	1.56	(0.77,3.18)	0.22	1.37	(0.65,2.86)	0.41	0.89	(0.59,1.34)	0.57	0.80	(0.61,1.06)	0.12	1.08	(0.75,1.56)	0.68	0.96	(0.67,1.38)	0.81

colnames(tabAll)[colnames(tabAll)=="allRows"]<-"predictor"

tabAll.PLWH.cd4<-tabAll[,c(1,11:13,17:19)]

#tabAll$cimt_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt_pvalue.unadj)))
#tabAll$cimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt_pvalue.adj)))
tabAll$clinevents_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$clinevents_pvalue.unadj)))
tabAll$clinevents_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$clinevents_pvalue.adj)))
tabAll$subclineventscimt_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$subclineventscimt_pvalue.unadj)))
tabAll$subclineventscimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$subclineventscimt_pvalue.adj)))
tabAll$clineventwithcimt_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$clineventwithcimt_pvalue.unadj)))
tabAll$clineventwithcimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$clineventwithcimt_pvalue.adj)))
write.csv(tabAll,row.names=F,file=paste(sep="_",outPrefix,"CIMT_CVDevents_PLWH.csv"))

10.2.1 Without CD4

rn.nocd4<-c("Sex: male","Age (in years)","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose","Cholesterol","CMV: persistently high titres","VZV: persistently high titres")

tabAll.nocd4 %>%
  dplyr::select(!allRows.nocd4) %>%
  kableRn(digits=3,escape=F,row_names=rn.nocd4,col.names=c(rep(c("RR","95% CI","p"),2),rep(c("RR","95% CI","p"),4)),caption="Model summaries for models for PLWH only 1) fitted to clinical CVD events only, 2) fitted to sub-clinical CIMT events only and 3) fitted to clinical CVD events and sub-clinical CIMT events (both unadjusted, single variable models and multiple regression models). The event count models are Poisson models and hence report rate ratios. Models fitted to multiply-imputed data from PLWH only.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  add_header_above(header=c(" "=1,"Unadjusted models" = 3, "Multiple regression model" = 3,"Unadjusted models" = 3, "Multiple regression model" = 3,"Unadjusted models" = 3, "Multiple regression model" = 3)) %>%
  add_header_above(header=c(" "=1,"Clinical CVD events" = 6, "Sub-clinical CIMT events" = 6,"Clinical CVD and sub-clinical CIMT events" = 6)) %>%
  kable_styling(full_width=F)

Table 10.7: Model summaries for models for PLWH only 1) fitted to clinical CVD events only, 2) fitted to sub-clinical CIMT events only and 3) fitted to clinical CVD events and sub-clinical CIMT events (both unadjusted, single variable models and multiple regression models). The event count models are Poisson models and hence report rate ratios. Models fitted to multiply-imputed data from PLWH only.
	Clinical CVD events						Sub-clinical CIMT events						Clinical CVD and sub-clinical CIMT events
	Unadjusted models			Multiple regression model			Unadjusted models			Multiple regression model			Unadjusted models			Multiple regression model
	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p
Sex: male	1.79	(0.36,8.80)	0.48	2.13	( 0.33,13.86)	0.43	0.48	(0.13,1.76)	0.27	0.17	(0.07,0.42)	0.00	0.85	(0.33,2.21)	0.74	0.66	(0.24,1.82)	0.42
Age (in years)	1.08	(1.02,1.15)	0.01	1.10	(1.02,1.19)	0.02	1.13	(1.09,1.17)	0.00	1.17	(1.13,1.22)	0.00	1.11	(1.07,1.14)	0.00	1.12	(1.07,1.18)	0.00
BMI	1.03	(0.90,1.18)	0.66	1.09	(0.94,1.26)	0.27	1.02	(0.89,1.17)	0.75	1.02	(0.92,1.13)	0.70	1.03	(0.93,1.13)	0.59	1.05	(0.94,1.17)	0.37
Diastolic blood pressure	1.01	(0.96,1.07)	0.61	1.02	(0.90,1.15)	0.76	1.04	(1.00,1.08)	0.04	1.06	(1.00,1.12)	0.06	1.03	(1.00,1.06)	0.07	1.03	(0.96,1.11)	0.41
Systolic blood pressure	1.01	(0.98,1.04)	0.38	0.99	(0.92,1.07)	0.82	1.03	(1.01,1.05)	0.01	1.00	(0.97,1.03)	0.89	1.02	(1.01,1.04)	0.01	1.00	(0.96,1.04)	0.85
Glucose	0.82	(0.37,1.78)	0.61	0.77	(0.35,1.71)	0.52	1.03	(0.80,1.33)	0.82	1.01	(0.85,1.21)	0.88	0.98	(0.74,1.31)	0.90	0.91	(0.64,1.30)	0.60
Cholesterol	0.83	(0.41,1.68)	0.61	0.74	(0.35,1.58)	0.44	1.23	(0.70,2.18)	0.47	0.69	(0.49,0.97)	0.03	1.04	(0.67,1.63)	0.86	0.73	(0.47,1.14)	0.17
CMV: persistently high titres	0.83	(0.11,6.34)	0.85	0.67	(0.08,6.01)	0.72	0.99	(0.21,4.75)	0.99	0.79	(0.31,2.02)	0.63	0.93	(0.27,3.19)	0.90	0.76	(0.22,2.60)	0.66
VZV: persistently high titres	0.03	(0.00, Inf)	1.00	0.03	(0.00, Inf)	1.00	7.95	( 1.87,33.89)	0.01	13.64	( 6.02,30.91)	0.00	2.96	(0.98,8.96)	0.06	3.28	(1.16,9.25)	0.03

colnames(tabAll.nocd4)[colnames(tabAll.nocd4)=="allRows"]<-"predictor"

tabAll.PLWH.nocd4<-tabAll.nocd4[,c(1,14:19)]

#tabAll$cimt_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt_pvalue.unadj)))
#tabAll$cimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll$cimt_pvalue.adj)))
tabAll.nocd4$clinevents_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll.nocd4$clinevents_pvalue.unadj)))
tabAll.nocd4$clinevents_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll.nocd4$clinevents_pvalue.adj)))
tabAll.nocd4$subclineventscimt_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll.nocd4$subclineventscimt_pvalue.unadj)))
tabAll.nocd4$subclineventscimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll.nocd4$subclineventscimt_pvalue.adj)))
tabAll.nocd4$clineventwithcimt_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll.nocd4$clineventwithcimt_pvalue.unadj)))
tabAll.nocd4$clineventwithcimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*>0",replace="0",tabAll.nocd4$clineventwithcimt_pvalue.adj)))

write.csv(tabAll.nocd4,row.names=F,file=paste(sep="_",outPrefix,"CIMT_CVDevents_PLWH.csv"))

10.2.1.1 Sensitivity analysis

Here we do a sensitivity analysis (for PLWH only) for the multivariable regression model (without CD4) showing an association between pVZV, respectively pCMV, and CVD events (clinical and sub-clinical CIMT).

To note that the main analysis made use of multiply imputed data and uses all available data to as best as possible impute the missing VZV status values, all the while taking into account the uncertainty inherent to the imputation itself.

For the sensitivity analysis we will manually impute all missing VZV statuses to “not persistently high titres” (scenario 1 - best case) and to “persistently high titres” (scenario 2 - worst case) respectively, and similarly for the sensitivity analysis for CMV status.

funFitAndSum_modClinEventsConfirmedCVDwithCIMT_SA<-function(varString,exp,scen,nDigits=2){
  tt<-complete(rhiccaDatWide_impMice.clinEventsConfirmed,include=TRUE,action="long")
  idxMVvzv<-which(is.na(tt$vzvStatus[tt$.imp==0]))
  idxMVcmv<-which(is.na(tt$cmvStatus[tt$.imp==0]))
  for(m in 1:max(tt$.imp)){
    if(scen=="worst"){
      if(exp=="vzv"){tt$vzvStatus[tt$.imp==m][idxMVvzv]<-"persistently high titres"}
      if(exp=="cmv"){tt$cmvStatus[tt$.imp==m][idxMVcmv]<-"persistently high titres"}
    }
    if(scen=="best"){
      if(exp=="vzv"){tt$vzvStatus[tt$.imp==m][idxMVvzv]<-"persistently low titres"}
      if(exp=="cmv"){tt$cmvStatus[tt$.imp==m][idxMVcmv]<-"persistently low titres"}
    }
  }
  tt<-as.mids(tt)
  
  mod<-with(tt %>% dplyr::filter(crf.ptype=="Patient"),glm(as.formula(paste(sep="","numClinEventsPerParticipantCVDwithCIMT~",varString)),offset=obsTime/(1000*365.25),family="quasipoisson"))
  
  sumObj<-summary(pool(mod)) %>%
    as.data.frame() %>%
    mutate(rownum=1:nrow(summary(pool(mod)))) %>%
    mutate(
      estimateLink=estimate,
      estimate=format(nsmall=nDigits,round(digits=nDigits,exp(estimate))),
      rrCI=NA
    ) %>%
    mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=T),cell_spec(format(nsmall=nDigits,round(digits=nDigits,p.value)),bold=F))) %>%
    mutate(p.value=gsub(p.value,pattern="0.000",replacement="&lt;0.001"))
  
  for(j in 1:nrow(sumObj)){
    sumObj$rrCI[j]<-paste(sep="","(",paste(collapse=",",format(nsmall=nDigits,round(digits=nDigits,c(exp(sumObj$estimateLink[j]-qnorm(0.975)*sumObj$std.error[j]),exp(sumObj$estimateLink[j]+qnorm(0.975)*sumObj$std.error[j]))))),")")
  }
  
  sumObj<-sumObj %>% dplyr::select(!estimateLink)
  
  return(sumObj)         
}

tab2.nocd4.sa.worst.vzv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT_SA("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)",scen="worst",exp="vzv")
tab2.nocd4.sa.worst.vzv$term<-as.character(tab2.nocd4.sa.worst.vzv$term)

tab2.nocd4.sa.best.vzv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT_SA("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)",scen="best",exp="vzv")
tab2.nocd4.sa.best.vzv$term<-as.character(tab2.nocd4.sa.best.vzv$term)

tab2.nocd4.sa.worst.cmv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT_SA("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)",scen="worst",exp="cmv")
tab2.nocd4.sa.worst.cmv$term<-as.character(tab2.nocd4.sa.worst.cmv$term)

tab2.nocd4.sa.best.cmv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT_SA("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)",scen="best",exp="cmv")
tab2.nocd4.sa.best.cmv$term<-as.character(tab2.nocd4.sa.best.cmv$term)

WORST case scenario - VZV

tab2.nocd4.sa.worst.vzv %>%
  kableRn(digits=3,escape=F,caption="WORST case scenario: every missing VZV status = persistently high VZV. The association with VZV is no longer statistically significant but this is hardly surprising as over half of the data will now be recorded as having pVZV.") %>%
  kable_styling()

Table 10.8: WORST case scenario: every missing VZV status = persistently high VZV. The association with VZV is no longer statistically significant but this is hardly surprising as over half of the data will now be recorded as having pVZV.
term	estimate	std.error	statistic	df	p.value	rownum	rrCI
(Intercept)	0.00	2.520	-4.136	664.291	0.00	1	(0.00,0.00)
crf.genderMale	0.68	0.491	-0.799	684.791	0.42	2	(0.26,1.77)
crf.age	1.12	0.023	4.940	687.357	0.00	3	(1.07,1.17)
physio.bmi_enrolment	1.05	0.051	0.991	638.375	0.32	4	(0.95,1.16)
physio.bp_dia_enrolment	1.04	0.036	1.121	674.430	0.26	5	(0.97,1.12)
physio.bp_sys_enrolment	0.99	0.019	-0.361	677.880	0.72	6	(0.96,1.03)
lims.Glucose_enrolment	0.90	0.187	-0.589	676.723	0.56	7	(0.62,1.29)
lims.Cholesterol_enrolment	0.74	0.223	-1.367	687.405	0.17	8	(0.48,1.14)
cmvStatus_2levelspersistently high titres	0.68	0.602	-0.634	148.979	0.53	9	(0.21,2.22)
vzvStatus_2levelspersistently high titres	3.08	0.498	2.263	113.974	0.03	10	(1.16,8.17)
lims.CD4_Absolute_enrolment	1.00	0.002	-1.620	689.392	0.11	11	(0.99,1.00)
lims.CD4_Absolute_changeFromBaselineAt24Months	1.00	0.002	0.232	156.559	0.82	12	(1.00,1.00)
log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)	0.96	0.185	-0.237	681.137	0.81	13	(0.67,1.38)

BEST case scenario - VZV

tab2.nocd4.sa.best.vzv %>%
  kableRn(digits=3,escape=F,caption="BEST case scenario: every missing VZV status = persistently low VZV. The association with VZV remains statistically significant - in fact becomes slightly stronger.") %>%
  kable_styling()

Table 10.9: BEST case scenario: every missing VZV status = persistently low VZV. The association with VZV remains statistically significant - in fact becomes slightly stronger.
term	estimate	std.error	statistic	df	p.value	rownum	rrCI
(Intercept)	0.00	2.520	-4.136	664.291	0.00	1	(0.00,0.00)
crf.genderMale	0.68	0.491	-0.799	684.791	0.42	2	(0.26,1.77)
crf.age	1.12	0.023	4.940	687.357	0.00	3	(1.07,1.17)
physio.bmi_enrolment	1.05	0.051	0.991	638.375	0.32	4	(0.95,1.16)
physio.bp_dia_enrolment	1.04	0.036	1.121	674.430	0.26	5	(0.97,1.12)
physio.bp_sys_enrolment	0.99	0.019	-0.361	677.880	0.72	6	(0.96,1.03)
lims.Glucose_enrolment	0.90	0.187	-0.589	676.723	0.56	7	(0.62,1.29)
lims.Cholesterol_enrolment	0.74	0.223	-1.367	687.405	0.17	8	(0.48,1.14)
cmvStatus_2levelspersistently high titres	0.68	0.602	-0.634	148.979	0.53	9	(0.21,2.22)
vzvStatus_2levelspersistently high titres	3.08	0.498	2.263	113.974	0.03	10	(1.16,8.17)
lims.CD4_Absolute_enrolment	1.00	0.002	-1.620	689.392	0.11	11	(0.99,1.00)
lims.CD4_Absolute_changeFromBaselineAt24Months	1.00	0.002	0.232	156.559	0.82	12	(1.00,1.00)
log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)	0.96	0.185	-0.237	681.137	0.81	13	(0.67,1.38)

WORST case scenario - CMV

tab2.nocd4.sa.worst.cmv %>%
  kableRn(digits=3,escape=F,caption="WORST case scenario: every missing CMV status = persistently high CMV. The association with CMV remains insignificant (and not that the association with VZV remains largely unchanged).") %>%
  kable_styling()

Table 10.10: WORST case scenario: every missing CMV status = persistently high CMV. The association with CMV remains insignificant (and not that the association with VZV remains largely unchanged).
term	estimate	std.error	statistic	df	p.value	rownum	rrCI
(Intercept)	0.00	2.520	-4.136	664.291	0.00	1	(0.00,0.00)
crf.genderMale	0.68	0.491	-0.799	684.791	0.42	2	(0.26,1.77)
crf.age	1.12	0.023	4.940	687.357	0.00	3	(1.07,1.17)
physio.bmi_enrolment	1.05	0.051	0.991	638.375	0.32	4	(0.95,1.16)
physio.bp_dia_enrolment	1.04	0.036	1.121	674.430	0.26	5	(0.97,1.12)
physio.bp_sys_enrolment	0.99	0.019	-0.361	677.880	0.72	6	(0.96,1.03)
lims.Glucose_enrolment	0.90	0.187	-0.589	676.723	0.56	7	(0.62,1.29)
lims.Cholesterol_enrolment	0.74	0.223	-1.367	687.405	0.17	8	(0.48,1.14)
cmvStatus_2levelspersistently high titres	0.68	0.602	-0.634	148.979	0.53	9	(0.21,2.22)
vzvStatus_2levelspersistently high titres	3.08	0.498	2.263	113.974	0.03	10	(1.16,8.17)
lims.CD4_Absolute_enrolment	1.00	0.002	-1.620	689.392	0.11	11	(0.99,1.00)
lims.CD4_Absolute_changeFromBaselineAt24Months	1.00	0.002	0.232	156.559	0.82	12	(1.00,1.00)
log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)	0.96	0.185	-0.237	681.137	0.81	13	(0.67,1.38)

BEST case scenario - CMV

tab2.nocd4.sa.best.cmv %>%
  kableRn(digits=3,escape=F,caption="BEST case scenario: every missing CMV status = persistently low CMV. The association with CMV remains insignificant (and not that the association with VZV remains largely unchanged).") %>%
  kable_styling()

Table 10.11: BEST case scenario: every missing CMV status = persistently low CMV. The association with CMV remains insignificant (and not that the association with VZV remains largely unchanged).
term	estimate	std.error	statistic	df	p.value	rownum	rrCI
(Intercept)	0.00	2.520	-4.136	664.291	0.00	1	(0.00,0.00)
crf.genderMale	0.68	0.491	-0.799	684.791	0.42	2	(0.26,1.77)
crf.age	1.12	0.023	4.940	687.357	0.00	3	(1.07,1.17)
physio.bmi_enrolment	1.05	0.051	0.991	638.375	0.32	4	(0.95,1.16)
physio.bp_dia_enrolment	1.04	0.036	1.121	674.430	0.26	5	(0.97,1.12)
physio.bp_sys_enrolment	0.99	0.019	-0.361	677.880	0.72	6	(0.96,1.03)
lims.Glucose_enrolment	0.90	0.187	-0.589	676.723	0.56	7	(0.62,1.29)
lims.Cholesterol_enrolment	0.74	0.223	-1.367	687.405	0.17	8	(0.48,1.14)
cmvStatus_2levelspersistently high titres	0.68	0.602	-0.634	148.979	0.53	9	(0.21,2.22)
vzvStatus_2levelspersistently high titres	3.08	0.498	2.263	113.974	0.03	10	(1.16,8.17)
lims.CD4_Absolute_enrolment	1.00	0.002	-1.620	689.392	0.11	11	(0.99,1.00)
lims.CD4_Absolute_changeFromBaselineAt24Months	1.00	0.002	0.232	156.559	0.82	12	(1.00,1.00)
log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)	0.96	0.185	-0.237	681.137	0.81	13	(0.67,1.38)

10.2.2 Table 3 for manuscript

10.2.2.1 CIMT

# rnames<-c("crf.ptypePatient",tabAll.PLWH.nocd4[,1],"lims.CD4_Absolute_enrolment","lims.CD4_Absolute_changeFromBaselineAt24Months")
# rn<-c("Participant type: PLWH","Sex: male","Age (years)","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","CMV: persistently high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (change from baseline at 24 months)")
# 
# colnames(tabAll.PLWH.nocd4)<-paste(sep="","PLWHnocd4_",colnames(tabAll.PLWH.nocd4))
# colnames(tabAll.PLWH.cd4)<-paste(sep="","PLWHcd4_",colnames(tabAll.PLWH.cd4))
# colnames(tabAll.all)<-paste(sep="","all_",colnames(tabAll.all))
# 
# tabAll.tab5manu<-cbind(
#   rnames,
#   tabAll.PLWH.cd4[match(rnames,tabAll.PLWH.cd4[,1]),2:4],
#   tabAll.PLWH.nocd4[match(rnames,tabAll.PLWH.nocd4[,1]),5:7],
#   tabAll.PLWH.cd4[match(rnames,tabAll.PLWH.cd4[,1]),5:7],
#   tabAll.all[match(rnames,tabAll.all[,1]),-1]
# )
# 
# tabAll.tab5manu %>%
#   dplyr::select(!"rnames") %>%
#   kableRn(digits=3,escape=F,row_names=rn,col.names=rep(c("RR","95% CI","p"),5),caption="Model summaries for models fitted to clinical CVD events and sub-clinical CIMT events (both unadjusted, single variable models and multiple regression models) for 1) PLWH and 2) all participants. The event count models are Poisson models and hence report rate ratios. Models fitted to multiply-imputed data.") %>%
#   #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
#   add_header_above(header=c(" "=1,"Unadjusted models" = 3, "Multiple regression model (without CD4)" = 3, "Multiple regression model (with CD4)" = 3,"Unadjusted models" = 3, "Multiple regression model" = 3)) %>%
#   add_header_above(header=c(" "=1,"PLWH" = 9,"All participants" = 6)) %>%
#   kable_styling(full_width=F)
# 
# tabAll.tab5manu$PLWHcd4_clineventwithcimt_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab5manu$PLWHcd4_clineventwithcimt_pvalue.unadj)))
# tabAll.tab5manu$PLWHnocd4_clineventwithcimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab5manu$PLWHnocd4_clineventwithcimt_pvalue.adj)))
# tabAll.tab5manu$PLWHcd4_clineventwithcimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab5manu$PLWHcd4_clineventwithcimt_pvalue.adj)))
# tabAll.tab5manu$all_clineventwithcimt_pvalue.unadj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab5manu$all_clineventwithcimt_pvalue.unadj)))
# tabAll.tab5manu$all_clineventwithcimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab5manu$all_clineventwithcimt_pvalue.adj)))
# 
# write.csv(tabAll.tab5manu,row.names=F,file=paste(sep="_",outPrefix,"CIMT_CVDevents_PLWH_All_Table3.csv"))

rnames<-c("crf.ptypePatient",tabAll.PLWH.nocd4[,1],"lims.CD4_Absolute_enrolment","lims.CD4_Absolute_changeFromBaselineAt24Months","log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")
rn<-c("Participant type: PLWH","Sex: male","Age (years)","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","CMV: persistently high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (change from baseline at 24 months)","log10 viral load (baseline)")

#colnames(tabAll.PLWH.nocd4)<-paste(sep="","PLWHnocd4_",colnames(tabAll.PLWH.nocd4))
colnames(tabAll.PLWH.cd4)<-paste(sep="","PLWHcd4_",colnames(tabAll.PLWH.cd4))
colnames(tabAll.all)<-paste(sep="","all_",colnames(tabAll.all))

tabAll.tab3manu<-cbind(
  rnames,
  #tabAll.PLWH.cd4[match(rnames,tabAll.PLWH.cd4[,1]),2:4],
  #tabAll.PLWH.nocd4[match(rnames,tabAll.PLWH.nocd4[,1]),5:7],
  tabAll.PLWH.cd4[match(rnames,tabAll.PLWH.cd4[,1]),-1],
  tabAll.all[match(rnames,tabAll.all[,1]),-1]
)

tabAll.tab3manu %>%
  dplyr::select(!"rnames") %>%
  kableRn(digits=3,escape=F,row_names=rn,col.names=rep(c("RR","95% CI","p"),4),caption="Model summaries for models fitted to clinical CVD events and sub-clinical CIMT events (both unadjusted, single variable models and multiple regression models) for 1) PLWH and 2) all participants. The event count models are quasi-Poisson models and hence report rate ratios. Models fitted to multiply-imputed data.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  add_header_above(header=c(" "=1,"Multiple regression model" = 3, "Multiple regression model" = 3,"Multiple regression model" = 3, "Multiple regression model" = 3)) %>%
  add_header_above(header=c(" "=1,"PLWH" = 6,"All participants" = 6)) %>%
  kable_styling(full_width=F)

Table 10.12: Model summaries for models fitted to clinical CVD events and sub-clinical CIMT events (both unadjusted, single variable models and multiple regression models) for 1) PLWH and 2) all participants. The event count models are quasi-Poisson models and hence report rate ratios. Models fitted to multiply-imputed data.
	PLWH						All participants
	Multiple regression model			Multiple regression model			Multiple regression model			Multiple regression model
	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p
Participant type: PLWH							3.73	( 1.37,10.16)	0.01	3.63	( 0.99,13.28)	0.05
Sex: male	0.16	(0.06,0.40)	0.00	0.68	(0.26,1.77)	0.42	0.33	(0.15,0.76)	0.01	0.87	(0.33,2.31)	0.78
Age (years)	1.18	(1.13,1.24)	0.00	1.12	(1.07,1.17)	0.00	1.18	(1.14,1.23)	0.00	1.14	(1.09,1.19)	0.00
BMI	1.04	(0.94,1.15)	0.45	1.05	(0.95,1.16)	0.32	0.96	(0.86,1.07)	0.43	1.03	(0.92,1.16)	0.58
Diastolic blood pressure	1.06	(1.00,1.13)	0.05	1.04	(0.97,1.12)	0.26	1.08	(1.02,1.14)	0.01	1.05	(0.98,1.12)	0.15
Systolic blood pressure	0.99	(0.96,1.02)	0.69	0.99	(0.96,1.03)	0.72	0.97	(0.95,1.00)	0.05	0.99	(0.95,1.02)	0.41
Glucose (baseline)	1.00	(0.83,1.21)	0.97	0.90	(0.62,1.29)	0.56	0.98	(0.80,1.21)	0.85	0.89	(0.62,1.29)	0.54
Cholesterol (baseline)	0.65	(0.45,0.94)	0.02	0.74	(0.48,1.14)	0.17	0.96	(0.71,1.29)	0.77	0.80	(0.54,1.20)	0.29
CMV: persistently high titres	0.87	(0.33,2.32)	0.78	0.68	(0.21,2.22)	0.53	0.87	(0.34,2.23)	0.77	0.72	(0.21,2.47)	0.60
VZV: persistently high titres	13.84	( 5.93,32.31)	0.00	3.08	(1.16,8.17)	0.03	6.38	( 3.08,13.23)	0.00	2.97	(1.15,7.67)	0.03
CD4 (baseline)	1.00	(1.00,1.00)	0.31	1.00	(0.99,1.00)	0.11
CD4 (change from baseline at 24 months)	1.00	(1.00,1.00)	0.54	1.00	(1.00,1.00)	0.82
log10 viral load (baseline)	0.80	(0.61,1.06)	0.12	0.96	(0.67,1.38)	0.81

tabAll.tab3manu$PLWHcd4_subclineventscimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$PLWHcd4_subclineventscimt_pvalue.adj)))
tabAll.tab3manu$PLWHcd4_clineventwithcimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$PLWHcd4_clineventwithcimt_pvalue.adj)))
tabAll.tab3manu$all_subclineventscimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$all_subclineventscimt_pvalue.adj)))
tabAll.tab3manu$all_clineventwithcimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$all_clineventwithcimt_pvalue.adj)))

write.csv(tabAll.tab3manu,row.names=F,file=paste(sep="_",outPrefix,"CIMT_SubAndClinCVDevents_PLWH_All_Table3.csv"))

10.2.2.2 CIMT (with ART)

# All participants
# tt<-complete(action="long",include=T,rhiccaDatWide_impMice.clinEventsConfirmed)
# tt<-tt %>%
#   dplyr::mutate(crf.artRegimen.2levels.bkp=crf.artRegimen.2levels) %>%
#   dplyr::mutate(
#       crf.artRegimen.2levels=case_when(
#       crf.ptype=="Patient"~crf.artRegimen.2levels.bkp,
#       crf.ptype=="Control"~levels(crf.artRegimen.2levels.bkp)[1]
#     )
#   )
# rhiccaDatWide_impMice.clinEventsConfirmed<-as.mids(tt)

## sub-clinical CIMT events only
tab1.ptype<-funFitAndSum_modClinEventsCIMTonly.all("crf.ptype")
tab1.sex<-funFitAndSum_modClinEventsCIMTonly.all("crf.gender")
tab1.age<-funFitAndSum_modClinEventsCIMTonly.all("crf.age")
tab1.smoke<-funFitAndSum_modClinEventsCIMTonly.all("crf.smoke") # ridiculous / unstable model coefficients due to no events in one group
tab1.alcohol<-funFitAndSum_modClinEventsCIMTonly.all("crf.alcohol") # ridiculous / unstable model coefficients due to no events in one group
tab1.waisthip<-funFitAndSum_modClinEventsCIMTonly.all("physio.waisthip_enrolment")
tab1.bmi<-funFitAndSum_modClinEventsCIMTonly.all("physio.bmi_enrolment")
tab1.diabp<-funFitAndSum_modClinEventsCIMTonly.all("physio.bp_dia_enrolment")
tab1.sysbp<-funFitAndSum_modClinEventsCIMTonly.all("physio.bp_sys_enrolment")
tab1.glucose<-funFitAndSum_modClinEventsCIMTonly.all("lims.Glucose_enrolment")
tab1.cholesterol<-funFitAndSum_modClinEventsCIMTonly.all("lims.Cholesterol_enrolment")
tab1.ckd<-funFitAndSum_modClinEventsCIMTonly.all("eGFR_existingCKD") # ridiculous / unstable model coefficients due to no events in one group
#tab1.cmv<-funFitAndSum_modClinEventsCIMTonly.all("cmvStatus") # ridiculous / unstable model coefficients due to no events in one group
#tab1.vzv<-funFitAndSum_modClinEventsCIMTonly.all("vzvStatus") # ridiculous / unstable model coefficients due to no events in one group
tab1.cmv<-funFitAndSum_modClinEventsCIMTonly.all("cmvStatus_2levels")
tab1.vzv<-funFitAndSum_modClinEventsCIMTonly.all("vzvStatus_2levels")

tab1<-rbind(tab1.ptype,
            tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv
)
tab1$term<-as.character(tab1$term)

tab2<-funFitAndSum_modClinEventsCIMTonly.all("crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels")
tab2$term<-as.character(tab2$term)

allRows<-setdiff(unique(c(tab1$term,tab2$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$rrCI.unadj<-tab1$rrCI[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.subclineventscimt<-tabAll

## clinical CVD events + sub-clinical CIMT events
tab1.ptype<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("crf.ptype")
tab1.sex<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("crf.gender")
tab1.age<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("crf.age")
tab1.smoke<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("crf.smoke") # ridiculous / unstable model coefficients due to no events in one group
tab1.alcohol<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("crf.alcohol") # ridiculous / unstable model coefficients due to no events in one group
tab1.waisthip<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("physio.waisthip_enrolment")
tab1.bmi<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("physio.bmi_enrolment")
tab1.diabp<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("physio.bp_dia_enrolment")
tab1.sysbp<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("physio.bp_sys_enrolment")
tab1.glucose<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("lims.Glucose_enrolment")
tab1.cholesterol<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("lims.Cholesterol_enrolment")
tab1.ckd<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("eGFR_existingCKD") # ridiculous / unstable model coefficients due to no events in one group
#tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("cmvStatus") # ridiculous / unstable model coefficients due to no events in one group
#tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("vzvStatus") # ridiculous / unstable model coefficients due to no events in one group
tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("cmvStatus_2levels")
tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("vzvStatus_2levels")

tab1<-rbind(tab1.ptype,
            tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv
)
tab1$term<-as.character(tab1$term)

tab2<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.all("crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels")
tab2$term<-as.character(tab2$term)

allRows<-setdiff(unique(c(tab1$term,tab2$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$rrCI.unadj<-tab1$rrCI[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.clineventswithcimt<-tabAll

## combining all in a single table
colnames(tab.subclineventscimt)<-paste(sep="_","subclineventscimt",colnames(tab.subclineventscimt))
colnames(tab.clineventswithcimt)<-paste(sep="_","clineventwithcimt",colnames(tab.clineventswithcimt))

allRows<-unique(c(tab.subclineventscimt$subclineventscimt_term,tab.clineventswithcimt$clineventswithcimt_term))
                
tabAll<-cbind(
  allRows,
  tab.subclineventscimt[match(allRows,tab.subclineventscimt$subclineventscimt_term),] %>% dplyr::select(!contains("term")),
  tab.clineventswithcimt[match(allRows,tab.clineventswithcimt$clineventwithcimt_term),] %>% dplyr::select(!contains("term"))
)

colnames(tabAll)[colnames(tabAll)=="allRows"]<-"predictor"

tabAll.all<-tabAll


# PLWH only

## sub-clinical CIMT events only
tab1.sex<-funFitAndSum_modClinEventsCIMTonly.hiv("crf.gender")
tab1.age<-funFitAndSum_modClinEventsCIMTonly.hiv("crf.age")
tab1.smoke<-funFitAndSum_modClinEventsCIMTonly.hiv("crf.smoke") # ridiculous / unstable model coefficients due to no events in one group
tab1.alcohol<-funFitAndSum_modClinEventsCIMTonly.hiv("crf.alcohol") # ridiculous / unstable model coefficients due to no events in one group
tab1.waisthip<-funFitAndSum_modClinEventsCIMTonly.hiv("physio.waisthip_enrolment")
tab1.bmi<-funFitAndSum_modClinEventsCIMTonly.hiv("physio.bmi_enrolment")
tab1.diabp<-funFitAndSum_modClinEventsCIMTonly.hiv("physio.bp_dia_enrolment")
tab1.sysbp<-funFitAndSum_modClinEventsCIMTonly.hiv("physio.bp_sys_enrolment")
tab1.glucose<-funFitAndSum_modClinEventsCIMTonly.hiv("lims.Glucose_enrolment")
tab1.cholesterol<-funFitAndSum_modClinEventsCIMTonly.hiv("lims.Cholesterol_enrolment")
tab1.ckd<-funFitAndSum_modClinEventsCIMTonly.hiv("eGFR_existingCKD") # ridiculous / unstable model coefficients due to no events in one group
#tab1.cmv<-funFitAndSum_modClinEventsCIMTonly.hiv("cmvStatus") # ridiculous / unstable model coefficients due to no events in one group
#tab1.vzv<-funFitAndSum_modClinEventsCIMTonly.hiv("vzvStatus") # ridiculous / unstable model coefficients due to no events in one group
tab1.cmv<-funFitAndSum_modClinEventsCIMTonly.hiv("cmvStatus_2levels")
tab1.vzv<-funFitAndSum_modClinEventsCIMTonly.hiv("vzvStatus_2levels")
tab1.cd4base<-funFitAndSum_modClinEventsCIMTonly.hiv("lims.CD4_Absolute_enrolment")
tab1.cd4change<-funFitAndSum_modClinEventsCIMTonly.hiv("lims.CD4_Absolute_changeFromBaselineAt24Months")
tab1.vl<-funFitAndSum_modClinEventsCIMTonly.hiv("log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")
tab1.art<-funFitAndSum_modClinEventsCIMTonly.hiv("crf.artRegimen.2levels")

tab1<-rbind(tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv,
            tab1.cd4base,
            tab1.cd4change,
            tab1.vl,
            tab1.art
)
tab1$term<-as.character(tab1$term)

tab2<-funFitAndSum_modClinEventsCIMTonly.hiv("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)+crf.artRegimen.2levels")
tab2$term<-as.character(tab2$term)

allRows<-setdiff(unique(c(tab1$term,tab2$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$rrCI.unadj<-tab1$rrCI[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.subclineventscimt<-tabAll


## clinical CVD events + sub-clinical CIMT events
tab1.sex<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("crf.gender")
tab1.age<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("crf.age")
tab1.smoke<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("crf.smoke") # ridiculous / unstable model coefficients due to no events in one group
tab1.alcohol<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("crf.alcohol") # ridiculous / unstable model coefficients due to no events in one group
tab1.waisthip<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("physio.waisthip_enrolment")
tab1.bmi<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("physio.bmi_enrolment")
tab1.diabp<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("physio.bp_dia_enrolment")
tab1.sysbp<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("physio.bp_sys_enrolment")
tab1.glucose<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("lims.Glucose_enrolment")
tab1.cholesterol<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("lims.Cholesterol_enrolment")
tab1.ckd<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("eGFR_existingCKD") # ridiculous / unstable model coefficients due to no events in one group
#tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("cmvStatus") # ridiculous / unstable model coefficients due to no events in one group
#tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("vzvStatus") # ridiculous / unstable model coefficients due to no events in one group
tab1.cmv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("cmvStatus_2levels")
tab1.vzv<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("vzvStatus_2levels")
tab1.cd4base<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("lims.CD4_Absolute_enrolment")
tab1.cd4change<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("lims.CD4_Absolute_changeFromBaselineAt24Months")
tab1.vl<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")
tab1.art<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("crf.artRegimen.2levels")


tab1<-rbind(tab1.sex,
            tab1.age,
            #tab1.smoke,
            #tab1.alcohol,
            tab1.bmi,
            tab1.diabp,
            tab1.sysbp,
            tab1.glucose,
            tab1.cholesterol,
            #tab1.ckd,
            tab1.cmv,
            tab1.vzv,
            tab1.cd4base,
            tab1.cd4change,
            tab1.vl,
            tab1.art
)
tab1$term<-as.character(tab1$term)

tab2<-funFitAndSum_modClinEventsConfirmedCVDwithCIMT.hiv("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)+crf.artRegimen.2levels")
tab2$term<-as.character(tab2$term)

allRows<-setdiff(unique(c(tab1$term,tab2$term)),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.unadj=NA,rrCI.unadj=NA,pvalue.unadj=NA,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.unadj<-tab1$estimate[match(tabAll$term,tab1$term)]
tabAll$rrCI.unadj<-tab1$rrCI[match(tabAll$term,tab1$term)]
tabAll$pvalue.unadj<-tab1$p.value[match(tabAll$term,tab1$term)]

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.clineventswithcimt<-tabAll


## combining all in a single table
colnames(tab.subclineventscimt)<-paste(sep="_","subclineventscimt",colnames(tab.subclineventscimt))
colnames(tab.clineventswithcimt)<-paste(sep="_","clineventwithcimt",colnames(tab.clineventswithcimt))

allRows<-unique(c(tab.subclineventscimt$subclineventscimt_term,tab.clineventswithcimt$clineventswithcimt_term))
                
tabAll<-cbind(
  allRows,
  tab.subclineventscimt[match(allRows,tab.subclineventscimt$subclineventscimt_term),] %>% dplyr::select(!contains("term")),
  tab.clineventswithcimt[match(allRows,tab.clineventswithcimt$clineventwithcimt_term),] %>% dplyr::select(!contains("term"))
)

colnames(tab.subclineventscimt.nocd4)<-paste(sep="_","subclineventscimt",colnames(tab.subclineventscimt.nocd4))
colnames(tab.clineventswithcimt.nocd4)<-paste(sep="_","clineventwithcimt",colnames(tab.clineventswithcimt.nocd4))

colnames(tabAll)[colnames(tabAll)=="allRows"]<-"predictor"

tabAll.PLWH.cd4<-tabAll

# Combining PLWH and all participants
rnames<-c("crf.ptypePatient",tabAll.PLWH.cd4[,1])
rn<-c("Participant type: PLWH","Sex: male","Age (years)","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","CMV: persistently high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (change from baseline at 24 months)","log10 viral load (baseline)","ART: no change to dolutegravir")

colnames(tabAll.PLWH.cd4)<-paste(sep="","PLWHcd4_",colnames(tabAll.PLWH.cd4))
colnames(tabAll.all)<-paste(sep="","all_",colnames(tabAll.all))

tabAll.tab3manu<-cbind(
  rnames,
  tabAll.PLWH.cd4[match(rnames,tabAll.PLWH.cd4[,1]),-c(1,2:4,8:10,14:16,20:22)],
  tabAll.all[match(rnames,tabAll.all[,1]),-c(1,2:4,8:10,14:16,20:22)]
)

tabAll.tab3manu %>%
  dplyr::select(!"rnames") %>%
  kableRn(digits=3,escape=F,row_names=rn,col.names=rep(c("RR","95% CI","p"),4),caption="Model summaries for models fitted to clinical CVD events and sub-clinical CIMT events (both unadjusted, single variable models and multiple regression models) for 1) PLWH and 2) all participants. The event count models are quasi-Poisson models and hence report rate ratios. Models fitted to multiply-imputed data.") %>%
  #pack_rows(index = c("Unadjusted models, all data" = 3, "Multiple regression model, all data" = 16, "Multiple regression model, PLWH only" = 18)) %>%
  add_header_above(header=c(" "=1,"Multiple regression model" = 3, "Multiple regression model" = 3,"Multiple regression model" = 3, "Multiple regression model" = 3)) %>%
  add_header_above(header=c(" "=1,"PLWH" = 6,"All participants" = 6)) %>%
  kable_styling(full_width=F)

Table 10.13: Model summaries for models fitted to clinical CVD events and sub-clinical CIMT events (both unadjusted, single variable models and multiple regression models) for 1) PLWH and 2) all participants. The event count models are quasi-Poisson models and hence report rate ratios. Models fitted to multiply-imputed data.
	PLWH						All participants
	Multiple regression model			Multiple regression model			Multiple regression model			Multiple regression model
	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p
Participant type: PLWH							3.73	( 1.37,10.16)	0.01	3.63	( 0.99,13.28)	0.05
Sex: male	0.14	(0.06,0.36)	0.00	0.68	(0.26,1.78)	0.43	0.33	(0.15,0.76)	0.01	0.87	(0.33,2.31)	0.78
Age (years)	1.19	(1.14,1.24)	0.00	1.12	(1.07,1.17)	0.00	1.18	(1.14,1.23)	0.00	1.14	(1.09,1.19)	0.00
BMI	1.04	(0.95,1.15)	0.38	1.05	(0.95,1.16)	0.32	0.96	(0.86,1.07)	0.43	1.03	(0.92,1.16)	0.58
Diastolic blood pressure	1.05	(1.00,1.12)	0.06	1.04	(0.97,1.12)	0.26	1.08	(1.02,1.14)	0.01	1.05	(0.98,1.12)	0.15
Systolic blood pressure	1.00	(0.97,1.03)	0.80	0.99	(0.96,1.03)	0.72	0.97	(0.95,1.00)	0.05	0.99	(0.95,1.02)	0.41
Glucose (baseline)	0.99	(0.83,1.18)	0.91	0.90	(0.62,1.29)	0.56	0.98	(0.80,1.21)	0.85	0.89	(0.62,1.29)	0.54
Cholesterol (baseline)	0.65	(0.45,0.93)	0.02	0.74	(0.48,1.14)	0.17	0.96	(0.71,1.29)	0.77	0.80	(0.54,1.20)	0.29
CMV: persistently high titres	0.92	(0.33,2.54)	0.88	0.68	(0.21,2.23)	0.53	0.87	(0.34,2.23)	0.77	0.72	(0.21,2.47)	0.60
VZV: persistently high titres	12.37	( 5.68,26.92)	0.00	3.09	(1.16,8.24)	0.03	6.38	( 3.08,13.23)	0.00	2.97	(1.15,7.67)	0.03
CD4 (baseline)	1.00	(1.00,1.00)	0.18	1.00	(0.99,1.00)	0.11
CD4 (change from baseline at 24 months)	1.00	(1.00,1.00)	0.32	1.00	(1.00,1.00)	0.82
log10 viral load (baseline)	0.75	(0.56,1.01)	0.06	0.96	(0.66,1.38)	0.82
ART: no change to dolutegravir	0.47	(0.22,1.01)	0.05	1.01	(0.43,2.34)	0.98

tabAll.tab3manu$PLWHcd4_subclineventscimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$PLWHcd4_subclineventscimt_pvalue.adj)))
tabAll.tab3manu$PLWHcd4_clineventwithcimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$PLWHcd4_clineventwithcimt_pvalue.adj)))
tabAll.tab3manu$all_subclineventscimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$all_subclineventscimt_pvalue.adj)))
tabAll.tab3manu$all_clineventwithcimt_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$all_clineventwithcimt_pvalue.adj)))

write.csv(tabAll.tab3manu,row.names=F,file=paste(sep="_",outPrefix,"CIMT_SubAndClinCVDevents_PLWH_All_TableS7.csv"))

10.2.2.3 PWV

# All participants

## PWV only
tab2<-funFitAndSum_modClinEventsPWVonly("crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels",filterPLWH=FALSE)
tab2$term<-as.character(tab2$term)

allRows<-setdiff(unique(tab2$term),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.subclineventspwv<-tabAll


## clinical events + PWV
tab2<-funFitAndSum_modClinEventsConfirmedCVDwithPWV("crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels",filterPLWH=FALSE)
tab2$term<-as.character(tab2$term)

allRows<-setdiff(unique(tab2$term),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.clineventswithpwv<-tabAll


## putting it all together
rm(tabAll)

#colnames(tab.clinevents)<-paste(sep="_","clinevents",colnames(tab.clinevents))
colnames(tab.subclineventspwv)<-paste(sep="_","subclineventspwv",colnames(tab.subclineventspwv))
colnames(tab.clineventswithpwv)<-paste(sep="_","clineventswithpwv",colnames(tab.clineventswithpwv))

allRows<-unique(c(tab.subclineventspwv$subclineventspwv_term,tab.clineventswithpwv$clineventswithpwv_term))

tabAll<-cbind(
  allRows,
  #tab.clinevents[match(allRows,tab.clinevents$clinevents_term),] %>% dplyr::select(!contains("term")),
  tab.subclineventspwv[match(allRows,tab.subclineventspwv$subclineventspwv_term),] %>% dplyr::select(!contains("term")),
  tab.clineventswithpwv[match(allRows,tab.clineventswithpwv$clineventswithpwv_term),] %>% dplyr::select(!contains("term"))
)

tabAll.all<-tabAll
rm(tabAll)

# PLWH

## PWV only
tab2<-funFitAndSum_modClinEventsPWVonly("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")
tab2$term<-as.character(tab2$term)

allRows<-setdiff(unique(tab2$term),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.subclineventspwv<-tabAll


## clinical events + PWV
tab2<-funFitAndSum_modClinEventsConfirmedCVDwithPWV("crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+cmvStatus_2levels+vzvStatus_2levels+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months+log10(lims.HIV_Abbott_Viral_Load_0pt2ml_P_enrolment)")
tab2$term<-as.character(tab2$term)

allRows<-setdiff(unique(tab2$term),"(Intercept)")

tabAll<-data.frame(term=allRows,rr.adj=NA,rrCI.adj=NA,pvalue.adj=NA)

tabAll$rr.adj<-tab2$estimate[match(tabAll$term,tab2$term)]
tabAll$rrCI.adj<-tab2$rrCI[match(tabAll$term,tab2$term)]
tabAll$pvalue.adj<-tab2$p.value[match(tabAll$term,tab2$term)]

tab.clineventswithpwv<-tabAll


## putting it all together
rm(tabAll)

#colnames(tab.clinevents)<-paste(sep="_","clinevents",colnames(tab.clinevents))
colnames(tab.subclineventspwv)<-paste(sep="_","subclineventspwv",colnames(tab.subclineventspwv))
colnames(tab.clineventswithpwv)<-paste(sep="_","clineventswithpwv",colnames(tab.clineventswithpwv))

allRows<-unique(c(tab.subclineventspwv$subclineventspwv_term,tab.clineventswithpwv$clineventswithpwv_term))

tabAll<-cbind(
  allRows,
  #tab.clinevents[match(allRows,tab.clinevents$clinevents_term),] %>% dplyr::select(!contains("term")),
  tab.subclineventspwv[match(allRows,tab.subclineventspwv$subclineventspwv_term),] %>% dplyr::select(!contains("term")),
  tab.clineventswithpwv[match(allRows,tab.clineventswithpwv$clineventswithpwv_term),] %>% dplyr::select(!contains("term"))
)

tabAll.PLWH<-tabAll
rm(tabAll)

# combining for the full table
rnames<-c(tabAll.all[,1],setdiff(tabAll.PLWH[,1],tabAll.all[,1]))
rn<-c("Participant type: PLWH","Sex: male","Age (years)","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","CMV: persistently high titres","VZV: persistently high titres","CD4 (baseline)","CD4 (change from baseline at 24 months)","log viral load (baseline)")

#colnames(tabAll.PLWH.nocd4)<-paste(sep="","PLWHnocd4_",colnames(tabAll.PLWH.nocd4))
colnames(tabAll.PLWH)<-paste(sep="","PLWH_",colnames(tabAll.PLWH))
colnames(tabAll.all)<-paste(sep="","all_",colnames(tabAll.all))

tabAll.tab3manu<-cbind(
  rnames,
  tabAll.PLWH[match(rnames,tabAll.PLWH[,1]),-1],
  tabAll.all[match(rnames,tabAll.all[,1]),-1]
)

tabAll.tab3manu %>%
  dplyr::select(!"rnames") %>%
  kableRn(digits=3,escape=F,row_names=rn,col.names=rep(c("RR","95% CI","p"),5),caption="Model summaries for multiple regression models fitted to clinical CVD events and sub-clinical PWV events for 1) PLWH and 2) all participants. The event count models are quasi-Poisson models and hence report rate ratios. Models fitted to multiply-imputed data.") %>%
  add_header_above(header=c(" "=1, "Multiple regression model" = 3, "Multiple regression model" = 3, "Multiple regression model"=3, "Multiple regression model"=3)) %>%
  add_header_above(header=c(" "=1,"PLWH" = 6,"All participants" = 6)) %>%
  kable_styling(full_width=F)

Table 10.14: Model summaries for multiple regression models fitted to clinical CVD events and sub-clinical PWV events for 1) PLWH and 2) all participants. The event count models are quasi-Poisson models and hence report rate ratios. Models fitted to multiply-imputed data.
	PLWH						All participants
	Multiple regression model			Multiple regression model			Multiple regression model			Multiple regression model
p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p	RR	95% CI	p
Participant type: PLWH							0.34	(0.13,0.88)	0.03	0.59	(0.25,1.38)	0.22
Sex: male	1.01	(0.27,3.73)	0.99	1.24	(0.43,3.58)	0.69	1.36	(0.55,3.37)	0.50	1.63	(0.72,3.71)	0.24
Age (years)	1.01	(0.95,1.08)	0.73	1.04	(1.00,1.09)	0.07	1.03	(0.99,1.08)	0.09	1.05	(1.02,1.09)	0.00
BMI	0.91	(0.75,1.11)	0.36	1.00	(0.88,1.13)	0.98	0.94	(0.85,1.03)	0.18	0.98	(0.90,1.07)	0.70
Diastolic blood pressure	1.00	(0.92,1.09)	0.97	1.01	(0.94,1.08)	0.73	1.04	(0.98,1.10)	0.22	1.03	(0.98,1.08)	0.29
Systolic blood pressure	1.03	(0.98,1.08)	0.22	1.01	(0.97,1.05)	0.50	1.00	(0.96,1.03)	0.87	1.00	(0.97,1.03)	0.98
Glucose (baseline)	1.16	(1.07,1.25)	0.00	1.12	(1.04,1.20)	0.00	1.14	(1.07,1.22)	0.00	1.10	(1.03,1.18)	0.00
Cholesterol (baseline)	1.15	(0.66,2.00)	0.62	0.95	(0.61,1.48)	0.83	1.34	(0.93,1.91)	0.11	1.09	(0.79,1.52)	0.59
CMV: persistently high titres	1.37	(0.36,5.17)	0.64	0.91	(0.29,2.85)	0.87	1.22	(0.37,3.99)	0.74	1.04	(0.36,2.98)	0.94
VZV: persistently high titres	1.89	(0.41,8.57)	0.41	1.27	(0.36,4.50)	0.72	1.46	(0.51,4.21)	0.48	1.33	(0.53,3.35)	0.54
CD4 (baseline)	1.00	(1.00,1.01)	0.29	1.00	(1.00,1.00)	0.92
CD4 (change from baseline at 24 months)	1.00	(1.00,1.00)	0.56	1.00	(1.00,1.00)	0.86
log viral load (baseline)	1.41	(0.79,2.51)	0.24	1.40	(0.87,2.24)	0.17

tabAll.tab3manu$PLWH_subclineventspwv_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$PLWH_subclineventspwv_pvalue.adj)))
tabAll.tab3manu$PLWH_clineventswithpwv_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$PLWH_clineventswithpwv_pvalue.adj)))
tabAll.tab3manu$all_subclineventspwv_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$all_subclineventspwv_pvalue.adj)))
tabAll.tab3manu$all_clineventswithpwv_pvalue.adj<-as.numeric(gsub(pattern="</span>",replacement="",gsub(pattern="<span.*0\\.",replace="0\\.",tabAll.tab3manu$all_clineventswithpwv_pvalue.adj)))

write.csv(tabAll.tab3manu,row.names=F,file=paste(sep="_",outPrefix,"PWV_SubAndClinCVDevents_PLWH_All_TableS5.csv"))

10.3 Time-to-event models

We can visualise clinical CVD events using the Kaplan-Meier estimator, but note that we observed too few events (9 in total) over the study period to make any meaningful analysis. Including sub-clinical CIMT events increases event numbers, but a time-to-event analysis is severely limited by the fact that CIMT sub-clinical events are only known up to a 24-month interval - the entire study period. As such it is more meaningful to consider the event count models given earlier.

if(!is.Date(rhiccaDatWide$lims.sample_date_enrolment)){rhiccaDatWide$lims.sample_date_enrolment<-as_date(mdy_hms(gsub(pattern=":[0-9][0-9][0-9]",replacement="",rhiccaDatWide$lims.sample_date_enrolment)))}
if(!is.Date(rhiccaDatWide$lims.sample_date_6months)){rhiccaDatWide$lims.sample_date_6months<-as_date(mdy_hms(gsub(pattern=":[0-9][0-9][0-9]",replacement="",rhiccaDatWide$lims.sample_date_6months)))}
if(!is.Date(rhiccaDatWide$lims.sample_date_12months)){rhiccaDatWide$lims.sample_date_12months<-as_date(mdy_hms(gsub(pattern=":[0-9][0-9][0-9]",replacement="",rhiccaDatWide$lims.sample_date_12months)))}
if(!is.Date(rhiccaDatWide$lims.sample_date_24months)){rhiccaDatWide$lims.sample_date_24months<-as_date(mdy_hms(gsub(pattern=":[0-9][0-9][0-9]",replacement="",rhiccaDatWide$lims.sample_date_24months)))}

clinEventsKM<-data.frame(
  pid=rhiccaDatWide$pid,
  ptype=rhiccaDatWide$crf.ptype,
  timeStart=NA,
  timeEvent=NA,
  status=0
) %>%
  filter(!is.na(pid))

clinEventsCvdKM<-clinEventsKM
clinEventsCvdWithPwvCimtAbpiKM<-clinEventsKM
clinEventsCvdWithPwvAbpiNoClinKM<-clinEventsKM
clinEventsSystKM<-clinEventsKM
clinEventsInfKM<-clinEventsKM

clinEventsSummary<-clinEventsDat %>%
  group_by(PID) %>%
  dplyr::summarise(
    timeFirstEvent=as.character(min(dmy(data_date),na.rm=T)),
    .groups="drop"
  ) %>%
  filter(!is.na(timeFirstEvent))

clinEventsCvdSummary<-clinEventsDat %>%
  dplyr::filter(category=="CVD") %>%
  group_by(PID) %>%
  dplyr::summarise(
    timeFirstEvent=as.character(min(dmy(data_date),na.rm=T)),
    .groups="drop"
  ) %>%
  filter(!is.na(timeFirstEvent))

clinEventsSystSummary<-clinEventsDat %>%
  dplyr::filter(category=="Systemic disease") %>%
  group_by(PID) %>%
  dplyr::summarise(
    timeFirstEvent=as.character(min(dmy(data_date),na.rm=T)),
    .groups="drop"
  ) %>%
  filter(!is.na(timeFirstEvent))

clinEventsInfSummary<-clinEventsDat %>%
  dplyr::filter(category=="Infection") %>%
  group_by(PID) %>%
  dplyr::summarise(
    timeFirstEvent=as.character(min(dmy(data_date),na.rm=T)),
    .groups="drop"
  ) %>%
  filter(!is.na(timeFirstEvent))

datTmp<-rhiccaDatLong %>%
  dplyr::filter(!is.na(pid)) %>%
  group_by(pid) %>%
  dplyr::summarise(
    maxDate=max(dmy(physio.data_date),na.rm=T),
    .groups="drop"
  )

clinEventsKM$timeStart<-dmy(rhiccaDatWide$crf.data_date[match(clinEventsKM$pid,rhiccaDatWide$pid)])
clinEventsKM$timeEvent<-ymd(clinEventsSummary$timeFirstEvent[match(clinEventsKM$pid,clinEventsSummary$PID)])
clinEventsKM$status[!is.na(as.character(clinEventsKM$timeEvent))]<-1
clinEventsKM$timeEvent[is.na(as.character(clinEventsKM$timeEvent))]<-datTmp$maxDate[match(clinEventsKM$pid[is.na(as.character(clinEventsKM$timeEvent))],datTmp$pid)]

# update start times for retrospectively recruited participants (see email from Laura 10/10/2021)
clinEventsKM$timeStart[clinEventsKM$pid=="RE1097G"]<-dmy("28jun2017")
clinEventsKM$timeStart[clinEventsKM$pid=="RE1316Q"]<-dmy("20jul2017")

datDeathTmp<-clinEventsDat %>%
  dplyr::filter(DIAGNOSIS=="Death" & data_date!="") %>%
  dplyr::select(PID,data_date)

clinEventsCvdKM$timeStart<-dmy(rhiccaDatWide$crf.data_date[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)])
clinEventsCvdKM$timeEvent<-ymd(clinEventsCvdSummary$timeFirstEvent[match(clinEventsCvdKM$pid,clinEventsCvdSummary$PID)])
clinEventsCvdKM$status[!is.na(as.character(clinEventsCvdKM$timeEvent))]<-1
clinEventsCvdKM$timeEvent[is.na(as.character(clinEventsCvdKM$timeEvent))]<-datTmp$maxDate[match(clinEventsCvdKM$pid[is.na(as.character(clinEventsCvdKM$timeEvent))],datTmp$pid)]
clinEventsCvdKM$timeEvent[clinEventsCvdKM$status==0 & is.element(el=clinEventsCvdKM$pid,set=datDeathTmp$PID)]<-dmy(datDeathTmp$data_date[match(clinEventsCvdKM$pid[clinEventsCvdKM$status==0 & is.element(el=clinEventsCvdKM$pid,set=datDeathTmp$PID)],datDeathTmp$PID)])
clinEventsCvdKM$timeStart[clinEventsCvdKM$pid=="RE1097G"]<-dmy("28jun2017")
clinEventsCvdKM$timeStart[clinEventsCvdKM$pid=="RE1316Q"]<-dmy("20jul2017")

clinEventsSystKM$timeStart<-dmy(rhiccaDatWide$crf.data_date[match(clinEventsSystKM$pid,rhiccaDatWide$pid)])
clinEventsSystKM$timeEvent<-ymd(clinEventsSystSummary$timeFirstEvent[match(clinEventsSystKM$pid,clinEventsSystSummary$PID)])
clinEventsSystKM$status[!is.na(as.character(clinEventsSystKM$timeEvent))]<-1
clinEventsSystKM$timeEvent[is.na(as.character(clinEventsSystKM$timeEvent))]<-datTmp$maxDate[match(clinEventsSystKM$pid[is.na(as.character(clinEventsSystKM$timeEvent))],datTmp$pid)]
clinEventsSystKM$timeEvent[clinEventsSystKM$status==0 & is.element(el=clinEventsSystKM$pid,set=datDeathTmp$PID)]<-dmy(datDeathTmp$data_date[match(clinEventsSystKM$pid[clinEventsSystKM$status==0 & is.element(el=clinEventsSystKM$pid,set=datDeathTmp$PID)],datDeathTmp$PID)])
clinEventsSystKM$timeStart[clinEventsSystKM$pid=="RE1097G"]<-dmy("28jun2017")
clinEventsSystKM$timeStart[clinEventsSystKM$pid=="RE1316Q"]<-dmy("20jul2017")

clinEventsInfKM$timeStart<-dmy(rhiccaDatWide$crf.data_date[match(clinEventsInfKM$pid,rhiccaDatWide$pid)])
clinEventsInfKM$timeEvent<-ymd(clinEventsInfSummary$timeFirstEvent[match(clinEventsInfKM$pid,clinEventsInfSummary$PID)])
clinEventsInfKM$status[!is.na(as.character(clinEventsInfKM$timeEvent))]<-1
clinEventsInfKM$timeEvent[is.na(as.character(clinEventsInfKM$timeEvent))]<-datTmp$maxDate[match(clinEventsInfKM$pid[is.na(as.character(clinEventsInfKM$timeEvent))],datTmp$pid)]
clinEventsInfKM$timeEvent[clinEventsInfKM$status==0 & is.element(el=clinEventsInfKM$pid,set=datDeathTmp$PID)]<-dmy(datDeathTmp$data_date[match(clinEventsInfKM$pid[clinEventsInfKM$status==0 & is.element(el=clinEventsInfKM$pid,set=datDeathTmp$PID)],datDeathTmp$PID)])
clinEventsInfKM$timeStart[clinEventsInfKM$pid=="RE1097G"]<-dmy("28jun2017")
clinEventsInfKM$timeStart[clinEventsInfKM$pid=="RE1316Q"]<-dmy("20jul2017")

clinEventsCvdWithPwvCimtAbpiKM<-clinEventsCvdKM
clinEventsCvdWithPwvCimtKM<-clinEventsCvdKM
clinEventsCvdWithCimtKM<-clinEventsCvdKM
clinEventsCvdWithPwvAbpiKM<-clinEventsCvdKM

clinEventsCvdPwvAbpiOnlyKM<-clinEventsCvdKM
clinEventsCvdPwvAbpiOnlyKM$timeEvent[clinEventsCvdPwvAbpiOnlyKM$status==1]<-datTmp$maxDate[match(clinEventsCvdPwvAbpiOnlyKM$pid[clinEventsCvdPwvAbpiOnlyKM$status==1],datTmp$pid)]
clinEventsCvdPwvAbpiOnlyKM$status<-0
clinEventsCvdPwvAbpiOnlyKM$survTime2<-Inf

clinEventsCvdWithPwvAbpiNoClinKM$timeStart<-dmy(rhiccaDatWide$crf.data_date[match(clinEventsCvdWithPwvAbpiNoClinKM$pid,rhiccaDatWide$pid)])
clinEventsCvdWithPwvAbpiNoClinKM$timeEvent<-datTmp$maxDate[match(clinEventsCvdWithPwvAbpiNoClinKM$pid,datTmp$pid)]

clinEventsNoDerivedKM<-clinEventsKM

# all of these are derived from lab variables only available at enrolment and 24 months...
# 44 additional events
for(pid in as.data.frame(rhiccaDatWide %>% filter(newOnsetDiabetes=="new diabetes" | newOnsetHypercholesterolemia=="new hypercholesterolemia" | newOnsetKidneyCondition=="new kidney condition") %>% dplyr::select(pid))[,1]){
  idx<-which(clinEventsKM$pid==pid)
  if(clinEventsKM$status[idx]==0){
    clinEventsKM$status[idx]<-3
    
    clinEventsKM$timeStart[idx]<-ymd(rhiccaDatWide$lims.sample_date_enrolment[idx])
    clinEventsKM$timeEvent[idx]<-ymd(rhiccaDatWide$lims.sample_date_24months[idx])
  }else if(clinEventsKM$status[idx]==3){
    # check that either no event is recorded yet OR (that this event occurred earlier AND that it has a shorter interval)
    if(is.na(clinEventsKM$timeStart[idx]) | (!is.na(clinEventsKM$timeStart[idx]) & clinEventsKM$timeStart[idx]>ymd(rhiccaDatWide$lims.sample_date_enrolment[idx]) & (ymd(rhiccaDatWide$lims.sample_date_24months[idx])-ymd(rhiccaDatWide$lims.sample_date_enrolment[idx]))<(clinEventsKM$timeEvent[idx]-clinEventsKM$timeStart[idx]))){
      clinEventsKM$timeStart[idx]<-ymd(rhiccaDatWide$lims.sample_date_enrolment[idx])
      clinEventsKM$timeEvent[idx]<-ymd(rhiccaDatWide$lims.sample_date_24months[idx])
    }
  }
}

for(pid in as.data.frame(rhiccaDatWide %>% filter(newOnsetDiabetes=="new diabetes" | newOnsetHypercholesterolemia=="new hypercholesterolemia" | newOnsetKidneyCondition=="new kidney condition") %>% dplyr::select(pid))[,1]){
  idx<-which(clinEventsSystKM$pid==pid)
  if(clinEventsSystKM$status[idx]==0){
    clinEventsSystKM$status[idx]<-3
    
    clinEventsSystKM$timeStart[idx]<-ymd(rhiccaDatWide$lims.sample_date_enrolment[idx])
    clinEventsSystKM$timeEvent[idx]<-ymd(rhiccaDatWide$lims.sample_date_24months[idx])
  }else if(clinEventsSystKM$status[idx]==3){
    # check that either no event is recorded yet OR (that this event occurred earlier AND that it has a shorter interval)
    if(is.na(clinEventsSystKM$timeStart[idx]) | (!is.na(clinEventsSystKM$timeStart[idx]) & clinEventsSystKM$timeStart[idx]>ymd(rhiccaDatWide$lims.sample_date_enrolment[idx]) & (ymd(rhiccaDatWide$lims.sample_date_24months[idx])-ymd(rhiccaDatWide$lims.sample_date_enrolment[idx]))<(clinEventsSystKM$timeEvent[idx]-clinEventsSystKM$timeStart[idx]))){
      clinEventsSystKM$timeStart[idx]<-ymd(rhiccaDatWide$lims.sample_date_enrolment[idx])
      clinEventsSystKM$timeEvent[idx]<-ymd(rhiccaDatWide$lims.sample_date_24months[idx])
    }
  }
}

for(pid in as.data.frame((rhiccaDatWide %>% filter(newOnsetCIMT=="new CIMT vasculopathy") %>% dplyr::select(pid)))[,1]){
  idx<-which(clinEventsCvdWithPwvCimtAbpiKM$pid==pid)
  if(clinEventsCvdWithPwvCimtAbpiKM$status[idx]==0){
    clinEventsCvdWithPwvCimtAbpiKM$status[idx]<-3
    
    clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]<-ymd(rhiccaDatWide$lims.sample_date_enrolment[idx])
    clinEventsCvdWithPwvCimtAbpiKM$timeEvent[idx]<-ymd(rhiccaDatWide$lims.sample_date_24months[idx])
  }else if(clinEventsCvdWithPwvCimtAbpiKM$status[idx]==3){
    # check that either no event is recorded yet OR (that this event occurred earlier AND that it has a shorter interval)
    if(is.na(clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]) | (!is.na(clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]) & clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]>ymd(rhiccaDatWide$lims.sample_date_enrolment[idx]) & (ymd(rhiccaDatWide$lims.sample_date_24months[idx])-ymd(rhiccaDatWide$lims.sample_date_enrolment[idx]))<(clinEventsCvdWithPwvCimtAbpiKM$timeEvent[idx]-clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]))){
      clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]<-ymd(rhiccaDatWide$lims.sample_date_enrolment[idx])
      clinEventsCvdWithPwvCimtAbpiKM$timeEvent[idx]<-ymd(rhiccaDatWide$lims.sample_date_24months[idx])
    }
  }
}

clinEventsCvdWithCimtKM<-clinEventsCvdWithPwvCimtAbpiKM

for(pid in as.data.frame((rhiccaDatWide %>% filter(newOnsetPWV=="new PWV vasculopathy") %>% dplyr::select(pid)))[,1]){
  idx<-which(clinEventsCvdWithPwvCimtAbpiKM$pid==pid)
  timePwv<-min(dmy(c(rhiccaDatWide$pwv.data_date_6months[idx][rhiccaDatWide$pwv.pwv_6months[idx]>12],rhiccaDatWide$pwv.data_date_12months[idx][rhiccaDatWide$pwv.pwv_12months[idx]>12],rhiccaDatWide$pwv.data_date_18months[idx][rhiccaDatWide$pwv.pwv_18months[idx]>12],rhiccaDatWide$pwv.data_date_24months[idx][rhiccaDatWide$pwv.pwv_24months[idx]>12],rhiccaDatWide$pwv.data_date_30months[idx][rhiccaDatWide$pwv.pwv_30months[idx]>12])),na.rm=T)
  
  if(length(timePwv)>0){
    timeStart<-dmy(rhiccaDatWide$pwv.data_date_enrolment[idx])
    timeEnd<-timePwv
  }
  
  if(clinEventsCvdWithPwvCimtAbpiKM$status[idx]==0){
    clinEventsCvdWithPwvCimtAbpiKM$status[idx]<-3
    clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]<-timeStart
    clinEventsCvdWithPwvCimtAbpiKM$timeEvent[idx]<-timeEnd
  }else if(clinEventsCvdWithPwvCimtAbpiKM$status[idx]==3){
    # check that either no event is recorded yet OR (that this event occurred earlier AND that it has a shorter interval)
    if(is.na(clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]) | (!is.na(clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]) & clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]>timeStart & (timeEnd-timeStart)<(clinEventsCvdWithPwvCimtAbpiKM$timeEvent[idx]-clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]))){
      clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]<-timeStart
      clinEventsCvdWithPwvCimtAbpiKM$timeEvent[idx]<-timeEnd
    }
  }
}

clinEventsCvdWithPwvCimtKM<-clinEventsCvdWithPwvCimtAbpiKM

for(pid in as.data.frame((rhiccaDatWide %>% filter(newOnsetABPI=="new ABPI vasculopathy") %>% dplyr::select(pid)))[,1]){
  idx<-which(clinEventsCvdWithPwvCimtAbpiKM$pid==pid)
  timeAbpi<-min(dmy(c(rhiccaDatWide$abpi.data_date_6months[idx][rhiccaDatWide$abpi.abpi_6months[idx]>0.9],rhiccaDatWide$abpi.data_date_12months[idx][rhiccaDatWide$abpi.abpi_12months[idx]>0.9],rhiccaDatWide$abpi.data_date_18months[idx][rhiccaDatWide$abpi.abpi_18months[idx]>0.9],rhiccaDatWide$abpi.data_date_24months[idx][rhiccaDatWide$abpi.abpi_14months[idx]>0.9],rhiccaDatWide$abpi.data_date_30months[idx][rhiccaDatWide$abpi.abpi_30months[idx]>0.9])),na.rm=T)
  
  if(length(timeAbpi)>0){
    timeStart<-dmy(rhiccaDatWide$abpi.data_date_enrolment[idx])
    timeEnd<-timeAbpi
  }
  
  if(clinEventsCvdWithPwvCimtAbpiKM$status[idx]==0){
    clinEventsCvdWithPwvCimtAbpiKM$status[idx]<-3
    clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]<-timeStart
    clinEventsCvdWithPwvCimtAbpiKM$timeEvent[idx]<-timeEnd
  }else if(clinEventsCvdWithPwvCimtAbpiKM$status[idx]==3){
    # check that either no event is recorded yet OR (that this event occurred earlier AND that it has a shorter interval)
    if(is.na(clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]) | (!is.na(clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]) & clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]>timeStart & (timeEnd-timeStart)<(clinEventsCvdWithPwvCimtAbpiKM$timeEvent[idx]-clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]))){
      clinEventsCvdWithPwvCimtAbpiKM$timeStart[idx]<-timeStart
      clinEventsCvdWithPwvCimtAbpiKM$timeEvent[idx]<-timeEnd
    }
  }
}

clinEventsCvdWithPwvCimtAbpiKM$timeStart[clinEventsCvdWithPwvCimtAbpiKM$pid=="RE1097G"]<-dmy("28jun2017")
clinEventsCvdWithPwvCimtAbpiKM$timeStart[clinEventsCvdWithPwvCimtAbpiKM$pid=="RE1316Q"]<-dmy("20jul2017")

clinEventsCvdWithPwvCimtKM$timeStart[clinEventsCvdWithPwvCimtKM$pid=="RE1097G"]<-dmy("28jun2017")
clinEventsCvdWithPwvCimtKM$timeStart[clinEventsCvdWithPwvCimtKM$pid=="RE1316Q"]<-dmy("20jul2017")

clinEventsCvdWithCimtKM$timeStart[clinEventsCvdWithCimtKM$pid=="RE1097G"]<-dmy("28jun2017")
clinEventsCvdWithCimtKM$timeStart[clinEventsCvdWithCimtKM$pid=="RE1316Q"]<-dmy("20jul2017")


for(pid in as.data.frame((rhiccaDatWide %>% filter(newOnsetPWV=="new PWV vasculopathy" | newOnsetABPI=="new ABPI vasculopathy") %>% dplyr::select(pid)))[,1]){
  idx<-which(clinEventsCvdWithPwvAbpiNoClinKM$pid==pid)
  timePwv<-min(dmy(c(rhiccaDatWide$pwv.data_date_6months[idx][rhiccaDatWide$pwv.pwv_6months[idx]>12],rhiccaDatWide$pwv.data_date_12months[idx][rhiccaDatWide$pwv.pwv_12months[idx]>12],rhiccaDatWide$pwv.data_date_18months[idx][rhiccaDatWide$pwv.pwv_18months[idx]>12],rhiccaDatWide$pwv.data_date_24months[idx][rhiccaDatWide$pwv.pwv_24months[idx]>12],rhiccaDatWide$pwv.data_date_30months[idx][rhiccaDatWide$pwv.pwv_30months[idx]>12])),na.rm=T)
  timeAbpi<-min(dmy(c(rhiccaDatWide$abpi.data_date_6months[idx][rhiccaDatWide$abpi.abpi_6months[idx]>0.9],rhiccaDatWide$abpi.data_date_12months[idx][rhiccaDatWide$abpi.abpi_12months[idx]>0.9],rhiccaDatWide$abpi.data_date_18months[idx][rhiccaDatWide$abpi.abpi_18months[idx]>0.9],rhiccaDatWide$abpi.data_date_24months[idx][rhiccaDatWide$abpi.abpi_14months[idx]>0.9],rhiccaDatWide$abpi.data_date_30months[idx][rhiccaDatWide$abpi.abpi_30months[idx]>0.9])),na.rm=T)
  
  if(length(timePwv)>0 & (length(timeAbpi)==0 | (length(timeAbpi)>0) & timePwv<=timeAbpi)){
      timeStart<-dmy(rhiccaDatWide$pwv.data_date_enrolment[idx])
      timeEnd<-timePwv
    }else if(length(timeAbpi)>0 & (length(timePwv)==0 | (length(timePwv)>0) & timeAbpi<=timePwv)){
      timeStart<-dmy(rhiccaDatWide$abpi.data_date_enrolment[idx])
      timeEnd<-timeAbpi
    }
  
  if(clinEventsCvdWithPwvAbpiNoClinKM$status[idx]==0){
    clinEventsCvdWithPwvAbpiNoClinKM$status[idx]<-3
    clinEventsCvdWithPwvAbpiNoClinKM$timeStart[idx]<-timeStart
    clinEventsCvdWithPwvAbpiNoClinKM$timeEvent[idx]<-timeEnd
  }else if(clinEventsCvdWithPwvAbpiNoClinKM$status[idx]==3){
    # check that either no event is recorded yet OR (that this event occurred earlier AND that it has a shorter interval)
    if(is.na(clinEventsCvdWithPwvAbpiNoClinKM$timeStart[idx]) | (!is.na(clinEventsCvdWithPwvAbpiNoClinKM$timeStart[idx]) & clinEventsCvdWithPwvAbpiNoClinKM$timeStart[idx]>timeStart & (timeEnd-timeStart)<(clinEventsCvdWithPwvAbpiNoClinKM$timeEvent[idx]-clinEventsCvdWithPwvAbpiNoClinKM$timeStart[idx]))){
      clinEventsCvdWithPwvAbpiNoClinKM$timeStart[idx]<-timeStart
      clinEventsCvdWithPwvAbpiNoClinKM$timeEvent[idx]<-timeEnd
    }
  }
}

clinEventsCvdWithPwvAbpiNoClinKM$timeStart[clinEventsCvdWithPwvAbpiNoClinKM$pid=="RE1097G"]<-dmy("28jun2017")
clinEventsCvdWithPwvAbpiNoClinKM$timeStart[clinEventsCvdWithPwvAbpiNoClinKM$pid=="RE1316Q"]<-dmy("20jul2017")

clinEventsKM<-clinEventsKM %>%
  mutate(
    timeEnrol=dmy(rhiccaDatWide$crf.data_date)
  ) %>%
  mutate(
    time=timeEvent-timeStart,
    survTime1=case_when(
      status==0~as.numeric(time),
      status==1~as.numeric(time),
      status==3~as.numeric(timeStart-timeEnrol)
    ),
    survTime2=case_when(
      status==0~Inf,
      status==1~as.numeric(time),
      status==3~as.numeric(timeEvent-timeEnrol)
    )
  )

clinEventsNoDerivedKM<-clinEventsNoDerivedKM %>%
  mutate(
    timeEnrol=dmy(rhiccaDatWide$crf.data_date)
  ) %>%
  mutate(
    time=timeEvent-timeStart,
    survTime1=case_when(
      status==0~as.numeric(time),
      status==1~as.numeric(time),
      status==3~as.numeric(timeStart-timeEnrol)
    ),
    survTime2=case_when(
      status==0~Inf,
      status==1~as.numeric(time),
      status==3~as.numeric(timeEvent-timeEnrol)
    )
  )

clinEventsCvdKM<-clinEventsCvdKM %>%
  mutate(
    timeEnrol=dmy(rhiccaDatWide$crf.data_date)
  ) %>%
  mutate(
    time=timeEvent-timeStart,
    survTime1=case_when(
      status==0~as.numeric(time),
      status==1~as.numeric(time),
      status==3~as.numeric(timeStart-timeEnrol)
    ),
    survTime2=case_when(
      status==0~Inf,
      status==1~as.numeric(time),
      status==3~as.numeric(timeEvent-timeEnrol)
    )
  )

clinEventsCvdKM.2years<-clinEventsCvdKM
maxTime<-2*365
for(j in 1:nrow(clinEventsCvdKM)){
  if(clinEventsCvdKM.2years$time[j]>maxTime){
    clinEventsCvdKM.2years$time[j]<-maxTime
    if(clinEventsCvdKM.2years$status[j]==0){
      clinEventsCvdKM.2years$survTime1[j]<-maxTime
    }else if(clinEventsCvdKM.2years$status[j]==1){
      clinEventsCvdKM.2years$survTime1[j]<-maxTime
      clinEventsCvdKM.2years$survTime2[j]<-Inf
      clinEventsCvdKM.2years$status[j]<-0
    }
  }
}

clinEventsSystKM<-clinEventsSystKM %>%
  mutate(
    timeEnrol=dmy(rhiccaDatWide$crf.data_date)
  ) %>%
  mutate(
    time=timeEvent-timeStart,
    survTime1=case_when(
      status==0~as.numeric(time),
      status==1~as.numeric(time),
      status==3~as.numeric(timeStart-timeEnrol)
    ),
    survTime2=case_when(
      status==0~Inf,
      status==1~as.numeric(time),
      status==3~as.numeric(timeEvent-timeEnrol)
    )
  )

clinEventsInfKM<-clinEventsInfKM %>%
  mutate(
    timeEnrol=dmy(rhiccaDatWide$crf.data_date)
  ) %>%
  mutate(
    time=timeEvent-timeStart,
    survTime1=case_when(
      status==0~as.numeric(time),
      status==1~as.numeric(time),
      status==3~as.numeric(timeStart-timeEnrol)
    ),
    survTime2=case_when(
      status==0~Inf,
      status==1~as.numeric(time),
      status==3~as.numeric(timeEvent-timeEnrol)
    )
  )

clinEventsCvdWithPwvCimtAbpiKM<-clinEventsCvdWithPwvCimtAbpiKM %>%
  mutate(
    timeEnrol=dmy(rhiccaDatWide$crf.data_date)
  ) %>%
  mutate(
    time=timeEvent-timeStart,
    survTime1=case_when(
      status==0~as.numeric(time),
      status==1~as.numeric(time),
      status==3~as.numeric(timeStart-timeEnrol)
    ),
    survTime2=case_when(
      status==0~Inf,
      status==1~as.numeric(time),
      status==3~as.numeric(timeEvent-timeEnrol)
    )
  )

clinEventsCvdWithPwvCimtKM<-clinEventsCvdWithPwvCimtKM %>%
  mutate(
    timeEnrol=dmy(rhiccaDatWide$crf.data_date)
  ) %>%
  mutate(
    time=timeEvent-timeStart,
    survTime1=case_when(
      status==0~as.numeric(time),
      status==1~as.numeric(time),
      status==3~as.numeric(timeStart-timeEnrol)
    ),
    survTime2=case_when(
      status==0~Inf,
      status==1~as.numeric(time),
      status==3~as.numeric(timeEvent-timeEnrol)
    )
  )

clinEventsCvdWithCimtKM<-clinEventsCvdWithCimtKM %>%
  mutate(
    timeEnrol=dmy(rhiccaDatWide$crf.data_date)
  ) %>%
  mutate(
    time=timeEvent-timeStart,
    survTime1=case_when(
      status==0~as.numeric(time),
      status==1~as.numeric(time),
      status==3~as.numeric(timeStart-timeEnrol)
    ),
    survTime2=case_when(
      status==0~Inf,
      status==1~as.numeric(time),
      status==3~as.numeric(timeEvent-timeEnrol)
    )
  )

clinEventsCvdWithPwvAbpiNoClinKM<-clinEventsCvdWithPwvAbpiNoClinKM %>%
  mutate(
    timeEnrol=dmy(rhiccaDatWide$crf.data_date)
  ) %>%
  mutate(
    time=timeEvent-timeStart,
    survTime1=case_when(
      status==0~as.numeric(time),
      status==1~as.numeric(time),
      status==3~as.numeric(timeStart-timeEnrol)
    ),
    survTime2=case_when(
      status==0~Inf,
      status==1~as.numeric(time),
      status==3~as.numeric(timeEvent-timeEnrol)
    )
  )


kmMod<-survfit(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsKM)
censMod<-icfit(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsKM)
censModLR<-ictest(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsKM)

censModAlt<-ic_np(cbind(survTime1,survTime2)~ptype,data=clinEventsKM) # different package, but otherwise same model as censMod

kmCvdMod<-               survfit(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdKM)
kmCvdMod.2years<-        survfit(Surv(time=survTime1, event=status, type="right") ~ ptype, data = clinEventsCvdKM.2years)
kmCvdWithCimtMod<-       survfit(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdWithCimtKM)
kmCvdWithPwvCimtMod<-    survfit(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdWithPwvCimtKM)
kmCvdWithPwvCimtAbpiMod<-survfit(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdWithPwvCimtAbpiKM)
kmCvdWithPwvAbpiNoClinMod<-survfit(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdWithPwvAbpiNoClinKM)


censCvdModLR<-ictest(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdKM)
censCvdModLR.2years<-ictest(Surv(time=survTime1, event=status, type="right") ~ ptype, data = clinEventsCvdKM.2years)
censCvdWithCimtModLR<-ictest(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdWithCimtKM)
censCvdWithPwvCimtMod<-icfit(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdWithPwvCimtKM)
censCvdWithPwvCimtModLR<-ictest(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdWithPwvCimtKM)
censCvdWithPwvCimtAbpiMod<-icfit(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdWithPwvCimtAbpiKM)
censCvdWithPwvCimtAbpiModLR<-ictest(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdWithPwvCimtAbpiKM)
censCvdWithPwvAbpiNoClinMod<-icfit(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdWithPwvAbpiNoClinKM)
censCvdWithPwvAbpiNoClinModLR<-ictest(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsCvdWithPwvAbpiNoClinKM)

#diag_baseline(censModAFT,dists="weibull")
#diag_covar(censModAFT)

logRankP<-censModLR$p.value
logRankPCvd<-censCvdModLR$p.value
logRankPCvd.2years<-censCvdModLR.2years$p.value
logRankPCvdWithCimt<-censCvdWithCimtModLR$p.value
logRankPCvdWithPwvCimt<-censCvdWithPwvCimtModLR$p.value
logRankPCvdWithPwvCimtAbpi<-censCvdWithPwvCimtAbpiModLR$p.value
logRankPCvdWithPwvAbpiNoClin<-censCvdWithPwvAbpiNoClinModLR$p.value

logRankNoIntCens<-survdiff(Surv(time, status) ~ ptype, data = clinEventsKM %>% filter(status!=3))
logRankPNoIntCens<-1-pchisq(logRankNoIntCens$chisq,df=1)

logRankNoIntCensCvd<-survdiff(Surv(time, status) ~ ptype, data = clinEventsCvdKM %>% filter(status!=3))
logRankPNoIntCensCvd<-1-pchisq(logRankNoIntCensCvd$chisq,df=1)

logRankNoIntCensCvd.2years<-survdiff(Surv(time, status) ~ ptype, data = clinEventsCvdKM.2years %>% filter(status!=3))
logRankPNoIntCensCvd.2years<-1-pchisq(logRankNoIntCensCvd.2years$chisq,df=1)

ggplot2:::autoplot(kmCvdMod) + 
 labs(x = "\n Time to first clinical event (days)", y = "Probability of no event\n", 
 title = "Kaplan-Meier curves for times until first clinical event by participant type.\nEvents include only recorded cardiovascular clinical events.",
 caption = paste(sep="","Log-rank test p-value p = ",format(nsmall=4,round(digits=4,logRankPNoIntCensCvd)))) +
 theme(plot.title = element_text(hjust = 0.5), 
 axis.title.x = element_text(face="bold", colour="#FF7A33", size = 12),
 axis.title.y = element_text(face="bold", colour="#FF7A33", size = 12),
 legend.title = element_text(face="bold", size = 10)) +
  theme(plot.title.position = "plot",plot.title = element_text(hjust = 0))

Figure 10.2: Kaplan-Meier curves for times until first CVD clinical event by participant type.

ggKM.nolegend<-autoplot(kmCvdMod.2years,) + 
  labs(x = "\n Time to first clinical CVD event (days)", y = "Probability of no event\n") +#, 
       #title = ,
       #caption = paste(sep="","Kaplan-Meier curves for times until first clinical event by participant type.\nEvents include only recorded cardiovascular clinical events.\nLog-rank test p-value p = ",format(nsmall=4,round(digits=4,logRankPNoIntCensCvd.2years)))) +
  theme(plot.title = element_text(hjust = 0.5), 
        axis.title.x = element_text(size = 12),
        axis.title.y = element_text(size = 12),
        legend.title = element_text(face="bold", size = 10)) +
  theme(plot.title.position = "plot",plot.title = element_text(hjust = 0)) +
  guides(fill="none",color="none") +
  scale_fill_manual(values=c("steelblue","orange"),labels=c("Control","PLWH"),name="") +
  scale_color_manual(values=c("steelblue","orange"),labels=c("Control","PLWH"),name="")

10.4 Multi-panel figure CIMT, PWV and CVD events

grid.arrange(nrow=1,gCIMTAdj.notitle,ggKM.nolegend)

# pdf(width=10,height=5,file=paste(sep="_",outPrefix,"CIMT_CVDevents_Figure2.pdf"))
# grid.arrange(nrow=1,gCIMTAdj.notitle,ggKM.nolegend)
# dev.off()

# grid.arrange(nrow=1,gCIMTAdj.notitle+ggtitle("A. Pre-clinical CVD (CIMT)"),gPWVAdj.notitle+ggtitle("B. Pre-clinical CVD (PWV)"),ggKM.nolegend+ggtitle("C. Clinical CVD events."))
# 
# pdf(width=15,height=6,file=paste(sep="_",outPrefix,"CIMT_PWV_CVDevents_SupFigure1.pdf"))
# grid.arrange(nrow=1,gCIMTAdj.notitle+ggtitle("A. Pre-clinical CVD (CIMT)"),gPWVAdj.notitle+ggtitle("B. Pre-clinical CVD (PWV)"),ggKM.nolegend+ggtitle("C. Clinical CVD events."))
# dev.off()

grid.arrange(nrow=1,gCIMTAdj.notitle+ggtitle("A. Pre-clinical CBD/CVD (CIMT)"),gPWVAdj.notitle+ggtitle("B. Pre-clinical CBD/CVD (PWV)"))

pdf(width=10,height=6,file=paste(sep="_",outPrefix,"CIMT_PWV_CVDevents_SupFigure1.pdf"))
grid.arrange(nrow=1,gCIMTAdj.notitle+ggtitle("A. Pre-clinical CBD/CVD (CIMT)"),gPWVAdj.notitle+ggtitle("B. Pre-clinical CBD/CVD (PWV)"))
dev.off()

## quartz_off_screen 
##                 2

11 Incidence rate of systemic events (mostly hypertension and hypercholesterolaemia) was prevalent in both groups.

From Table 9.1, we see that systemic events are more prevalent in the control group than in the PLWH group (~ double the event rate in controls compared to PLWH). We can drill a bit deeper into this and break down the systemic events into different categories as done in Table 11.1.

clinEventsSystemic<-data.frame(
  pid="PID",
  crf.PLWH="PTYPE",
  type="TYPE"
)

for(j in which(rhiccaDatWide$clinEventCategoryConfirmedSystemicType!="")){
  evts<-unlist(strsplit(split=";",rhiccaDatWide$clinEventCategoryConfirmedSystemicType[j]))
  for(evt in evts){
    clinEventsSystemic<-rbind(clinEventsSystemic,c(rhiccaDatWide$pid[j],as.character(rhiccaDatWide$crf.PLWH[j]),evt))
  }
}
clinEventsSystemic<-clinEventsSystemic[-1,]
clinEventsSystemic$crf.PLWH<-factor(clinEventsSystemic$crf.PLWH,levels=c("PLWH","Control"))
clinEventsSystemic$type[clinEventsSystemic$typ=="chronic renal failure"]<-"kidney disease"

tt<-table(clinEventsSystemic$type,clinEventsSystemic$crf.PLWH)
tt<-data.frame(
  All=tt[,"PLWH"]+tt[,"Control"],
  PLWH=tt[,"PLWH"],
  Control=tt[,"Control"]
)
tt<-rbind(colSums(tt),tt)
rownames(tt)[1]<-"Any"
tt<-tt[c("Any","death","hypercholesterolemia","hypertension","diabetes","kidney disease","anaemia"),]
rownames(tt)<-c("Any","Death","Hypercholesterolemia","Hypertension","Diabetes","Kidney disease","Anaemia")

ttCount<-tt
for(j in 1:ncol(ttCount)){
  if(colnames(tt)[j]!="All"){
    tt[,j]<-tt[,j]/sum(as.numeric(rhiccaDatWide$obsTime[rhiccaDatWide$crf.PLWH==colnames(tt)[j]])/(1000*365.25))
  }else{
    tt[,j]<-tt[,j]/sum(as.numeric(rhiccaDatWide$obsTime)/(1000*365.25))
  }
}

ttCountCount<-ttCount

for(j in 1:ncol(ttCount)){
  for(i in 1:nrow(ttCount)){
    ttCount[i,j]<-paste(sep="",ttCount[i,j]," (",format(nsmall=3,round(digits=3,tt[i,j])),")")
  }
}

ttCount %>%
  kable(escape=F,caption="Listing of systemic disease event types. The numbers in brackets is the rate of events per 1000 person-years.") %>%
  kable_styling(full_width = F)

Table 11.1: Listing of systemic disease event types. The numbers in brackets is the rate of events per 1000 person-years.
	All	PLWH	Control
Any	89 (51.933)	57 (43.736)	32 (77.958)
Death	3 (1.751)	3 (2.302)	0 (0.000)
Hypercholesterolemia	41 (23.924)	25 (19.183)	16 (38.979)
Hypertension	20 (11.670)	11 (8.440)	9 (21.926)
Diabetes	3 (1.751)	1 (0.767)	2 (4.872)
Kidney disease	21 (12.254)	16 (12.277)	5 (12.181)
Anaemia	1 (0.584)	1 (0.767)	0 (0.000)

Looking at an event count model, we get Table 11.2.

modClinEventsConfirmedSystemic<-with(rhiccaDatWide_impMice.clinEventsConfirmed,glm(numClinEventsPerParticipantSystemic~crf.ptype+crf.gender+crf.age+physio.bmi_enrolment+cmvStatus+vzvStatus,offset=obsTime/(1000*365.25),family="quasipoisson"))

summary(pool(modClinEventsConfirmedSystemic)) %>%
  as.data.frame() %>%
  dplyr::select(!term) %>%
  mutate(rownum=1:nrow(summary(pool(modClinEventsConfirmedSystemic)))) %>%
  mutate(estimate=exp(estimate)) %>%
  mutate(p.value=ifelse(rownum>1 & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  dplyr::select(!c(rownum)) %>%
  kableRn(caption="Results for a Poisson regression model for the number of confirmed systemic clinical events per 1,000 person-years, including new onset diabetes and hypercholesterolemia events, with predictors participant type, age, sex, waist-hip ratio, diastolic and systolic blood pressure, glucose and cholesterol (all baseline measurements). Model fitted to multiply imputed data.",digits=4,row_names=c("Intercept","Participant type: PLWH","Sex male","Age (in years)","BMI","CMV: progression to latency","CMV: reactivated latent infection","CMV: persistently high titres","VZV: progression to latency","VZV: reactivated latent infection","VZV: persistently high titres"),escape=F) %>%
  kable_styling(full_width=F)

Table 11.2: Results for a Poisson regression model for the number of confirmed systemic clinical events per 1,000 person-years, including new onset diabetes and hypercholesterolemia events, with predictors participant type, age, sex, waist-hip ratio, diastolic and systolic blood pressure, glucose and cholesterol (all baseline measurements). Model fitted to multiply imputed data.
	estimate	std.error	statistic	df	p.value
Intercept	0.0582	0.9577	-2.9689	873.2665	0.0031
Participant type: PLWH	0.3980	0.2836	-3.2479	855.7700	0.0012
Sex male	0.5199	0.2445	-2.6751	865.4556	0.0076
Age (in years)	1.0301	0.0116	2.5496	861.5986	0.0110
BMI	0.9950	0.0251	-0.2011	877.9517	0.8407
CMV: progression to latency	1.3793	0.9333	0.3446	143.4293	0.7309
CMV: reactivated latent infection	2.1333	0.3426	2.2112	310.3312	0.0278
CMV: persistently high titres	1.8481	0.3206	1.9157	729.5477	0.0558
VZV: progression to latency	1.1165	0.4104	0.2686	208.5760	0.7885
VZV: reactivated latent infection	1.2362	0.6772	0.3131	223.7142	0.7545
VZV: persistently high titres	1.1780	0.3331	0.4918	278.0674	0.6232

This trend, HIV being protective, holds up in a time-to-event model.

clinEventsKM<-clinEventsKM %>%
  mutate(
    sex=rhiccaDatWide$crf.gender[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    age=rhiccaDatWide$crf.age[match(clinEventsKM$pid,rhiccaDatWide$pid)]-35, # centering on 35 years
    waisthip=rhiccaDatWide$physio.waisthip_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    bmi=rhiccaDatWide$physio.bmi_cat_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    diaBp=rhiccaDatWide$physio.bp_dia_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    sysBp=rhiccaDatWide$physio.bp_sys_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    cimtBase=rhiccaDatWide$cimt.cimt_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    cimtChange=(rhiccaDatWide$cimt.cimt_24months-rhiccaDatWide$cimt.cimt_enrolment)[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    pwvBase=rhiccaDatWide$pwv.pwv_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    pwvChange=(rhiccaDatWide$pwv.pwv_24months-rhiccaDatWide$pwv.pwv_enrolment)[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    glucose=rhiccaDatWide$lims.Glucose_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    cholesterol=rhiccaDatWide$lims.Cholesterol_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
  ) %>%
  mutate(
    waisthipcat=factor(case_when(
      sex=="Female" & waisthip<=0.8 ~ "low",
      sex=="Female" & waisthip>0.8 & waisthip<=0.85 ~ "moderate",
      sex=="Female" & waisthip>0.85 ~ "high",
      sex=="Male" & waisthip<=0.95 ~ "low",
      sex=="Male" & waisthip>0.95 & waisthip<=1 ~ "moderate",
      sex=="Male" & waisthip>1 ~ "high"
    ),levels=c("moderate","low","high"))
  )

clinEventsNoDerivedKM<-clinEventsNoDerivedKM %>%
  mutate(
    sex=rhiccaDatWide$crf.gender[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    age=rhiccaDatWide$crf.age[match(clinEventsKM$pid,rhiccaDatWide$pid)]-35, # centering on 35 years
    waisthip=rhiccaDatWide$physio.waisthip_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    bmi=rhiccaDatWide$physio.bmi_cat_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    diaBp=rhiccaDatWide$physio.bp_dia_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    sysBp=rhiccaDatWide$physio.bp_sys_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    cimtBase=rhiccaDatWide$cimt.cimt_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    cimtChange=(rhiccaDatWide$cimt.cimt_24months-rhiccaDatWide$cimt.cimt_enrolment)[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    pwvBase=rhiccaDatWide$pwv.pwv_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    pwvChange=(rhiccaDatWide$pwv.pwv_24months-rhiccaDatWide$pwv.pwv_enrolment)[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    glucose=rhiccaDatWide$lims.Glucose_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
    cholesterol=rhiccaDatWide$lims.Cholesterol_enrolment[match(clinEventsKM$pid,rhiccaDatWide$pid)],
  ) %>%
  mutate(
    waisthipcat=factor(case_when(
      sex=="Female" & waisthip<=0.8 ~ "low",
      sex=="Female" & waisthip>0.8 & waisthip<=0.85 ~ "moderate",
      sex=="Female" & waisthip>0.85 ~ "high",
      sex=="Male" & waisthip<=0.95 ~ "low",
      sex=="Male" & waisthip>0.95 & waisthip<=1 ~ "moderate",
      sex=="Male" & waisthip>1 ~ "high"
    ),levels=c("moderate","low","high"))
  )

clinEventsCvdKM<-clinEventsCvdKM %>%
  mutate(
    sex=rhiccaDatWide$crf.gender[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)],
    age=rhiccaDatWide$crf.age[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)]-35, # centering on 35 years
    waisthip=rhiccaDatWide$physio.waisthip_enrolment[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)],
    bmi=rhiccaDatWide$physio.bmi_cat_enrolment[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)],
    diaBp=rhiccaDatWide$physio.bp_dia_enrolment[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)],
    sysBp=rhiccaDatWide$physio.bp_sys_enrolment[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)],
    cimtBase=rhiccaDatWide$cimt.cimt_enrolment[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)],
    cimtChange=(rhiccaDatWide$cimt.cimt_24months-rhiccaDatWide$cimt.cimt_enrolment)[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)],
    pwvBase=rhiccaDatWide$pwv.pwv_enrolment[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)],
    pwvChange=(rhiccaDatWide$pwv.pwv_24months-rhiccaDatWide$pwv.pwv_enrolment)[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)],
    glucose=rhiccaDatWide$lims.Glucose_enrolment[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)],
    cholesterol=rhiccaDatWide$lims.Cholesterol_enrolment[match(clinEventsCvdKM$pid,rhiccaDatWide$pid)],
  ) %>%
  mutate(
    waisthipcat=factor(case_when(
      sex=="Female" & waisthip<=0.8 ~ "low",
      sex=="Female" & waisthip>0.8 & waisthip<=0.85 ~ "moderate",
      sex=="Female" & waisthip>0.85 ~ "high",
      sex=="Male" & waisthip<=0.95 ~ "low",
      sex=="Male" & waisthip>0.95 & waisthip<=1 ~ "moderate",
      sex=="Male" & waisthip>1 ~ "high"
    ),levels=c("moderate","low","high"))
  )

clinEventsCvdKM.2years<-clinEventsCvdKM
maxTime<-2*365
for(j in 1:nrow(clinEventsCvdKM)){
  if(clinEventsCvdKM.2years$time[j]>maxTime){
    clinEventsCvdKM.2years$time[j]<-maxTime
    if(clinEventsCvdKM.2years$status[j]==0){
      clinEventsCvdKM.2years$survTime1[j]<-maxTime
    }else if(clinEventsCvdKM.2years$status[j]==1){
      clinEventsCvdKM.2years$survTime1[j]<-maxTime
      clinEventsCvdKM.2years$survTime2[j]<-Inf
      clinEventsCvdKM.2years$status[j]<-0
    }
  }
}

clinEventsSystKM<-clinEventsSystKM %>%
  mutate(
    sex=rhiccaDatWide$crf.gender[match(clinEventsSystKM$pid,rhiccaDatWide$pid)],
    age=rhiccaDatWide$crf.age[match(clinEventsSystKM$pid,rhiccaDatWide$pid)]-35, # centering on 35 years
    waisthip=rhiccaDatWide$physio.waisthip_enrolment[match(clinEventsSystKM$pid,rhiccaDatWide$pid)],
    bmi=rhiccaDatWide$physio.bmi_cat_enrolment[match(clinEventsSystKM$pid,rhiccaDatWide$pid)],
    diaBp=rhiccaDatWide$physio.bp_dia_enrolment[match(clinEventsSystKM$pid,rhiccaDatWide$pid)],
    sysBp=rhiccaDatWide$physio.bp_sys_enrolment[match(clinEventsSystKM$pid,rhiccaDatWide$pid)],
    cimtBase=rhiccaDatWide$cimt.cimt_enrolment[match(clinEventsSystKM$pid,rhiccaDatWide$pid)],
    cimtChange=(rhiccaDatWide$cimt.cimt_24months-rhiccaDatWide$cimt.cimt_enrolment)[match(clinEventsSystKM$pid,rhiccaDatWide$pid)],
    pwvBase=rhiccaDatWide$pwv.pwv_enrolment[match(clinEventsSystKM$pid,rhiccaDatWide$pid)],
    pwvChange=(rhiccaDatWide$pwv.pwv_24months-rhiccaDatWide$pwv.pwv_enrolment)[match(clinEventsSystKM$pid,rhiccaDatWide$pid)],
    glucose=rhiccaDatWide$lims.Glucose_enrolment[match(clinEventsSystKM$pid,rhiccaDatWide$pid)],
    cholesterol=rhiccaDatWide$lims.Cholesterol_enrolment[match(clinEventsSystKM$pid,rhiccaDatWide$pid)],
  ) %>%
  mutate(
    waisthipcat=factor(case_when(
      sex=="Female" & waisthip<=0.8 ~ "low",
      sex=="Female" & waisthip>0.8 & waisthip<=0.85 ~ "moderate",
      sex=="Female" & waisthip>0.85 ~ "high",
      sex=="Male" & waisthip<=0.95 ~ "low",
      sex=="Male" & waisthip>0.95 & waisthip<=1 ~ "moderate",
      sex=="Male" & waisthip>1 ~ "high"
    ),levels=c("moderate","low","high"))
  )

clinEventsInfKM<-clinEventsInfKM %>%
  mutate(
    sex=rhiccaDatWide$crf.gender[match(clinEventsInfKM$pid,rhiccaDatWide$pid)],
    age=rhiccaDatWide$crf.age[match(clinEventsInfKM$pid,rhiccaDatWide$pid)]-35, # centering on 35 years
    waisthip=rhiccaDatWide$physio.waisthip_enrolment[match(clinEventsInfKM$pid,rhiccaDatWide$pid)],
    bmi=rhiccaDatWide$physio.bmi_cat_enrolment[match(clinEventsCvdWithPwvCimtAbpiKM$pid,rhiccaDatWide$pid)],
    diaBp=rhiccaDatWide$physio.bp_dia_enrolment[match(clinEventsInfKM$pid,rhiccaDatWide$pid)],
    sysBp=rhiccaDatWide$physio.bp_sys_enrolment[match(clinEventsInfKM$pid,rhiccaDatWide$pid)],
    cimtBase=rhiccaDatWide$cimt.cimt_enrolment[match(clinEventsInfKM$pid,rhiccaDatWide$pid)],
    cimtChange=(rhiccaDatWide$cimt.cimt_24months-rhiccaDatWide$cimt.cimt_enrolment)[match(clinEventsInfKM$pid,rhiccaDatWide$pid)],
    pwvBase=rhiccaDatWide$pwv.pwv_enrolment[match(clinEventsInfKM$pid,rhiccaDatWide$pid)],
    pwvChange=(rhiccaDatWide$pwv.pwv_24months-rhiccaDatWide$pwv.pwv_enrolment)[match(clinEventsInfKM$pid,rhiccaDatWide$pid)],
    glucose=rhiccaDatWide$lims.Glucose_enrolment[match(clinEventsInfKM$pid,rhiccaDatWide$pid)],
    cholesterol=rhiccaDatWide$lims.Cholesterol_enrolment[match(clinEventsInfKM$pid,rhiccaDatWide$pid)],
  ) %>%
  mutate(
    waisthipcat=factor(case_when(
      sex=="Female" & waisthip<=0.8 ~ "low",
      sex=="Female" & waisthip>0.8 & waisthip<=0.85 ~ "moderate",
      sex=="Female" & waisthip>0.85 ~ "high",
      sex=="Male" & waisthip<=0.95 ~ "low",
      sex=="Male" & waisthip>0.95 & waisthip<=1 ~ "moderate",
      sex=="Male" & waisthip>1 ~ "high"
    ),levels=c("moderate","low","high"))
  )

censModSystAFT<-ic_par(cbind(survTime1,survTime2)~ptype+sex+age+waisthipcat,data=clinEventsSystKM,model = "aft",dist = "weibull")

summary(censModSystAFT)$summaryParameters %>%
  as.data.frame() %>%
  mutate(rownum=1:nrow(summary(censModSystAFT)$summaryParameters)) %>%
  mutate(p=ifelse(rownum>2 & p<0.05,cell_spec(format(nsmall=4,round(digits=4,p)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p)),bold=F))) %>%
  dplyr::select(!rownum) %>%
  kableRn(row_names = c("log Weibull shape parameter","log Weibull scale parameter","Participant type: PLWH","Sex: male","Age (years)","Waist-hip ratio (low)","Waist-hip ratio (high)"),escape=FALSE,caption="Results from an AFT model for clinical systemic events only.") %>%
  kable_styling(full_width = F)

Table 11.3: Results from an AFT model for clinical systemic events only.
	Estimate	Exp(Est)	Std.Error	z-value	p
log Weibull shape parameter	-0.3979	6.718e-01	0.16920	-2.3510	0.0187
log Weibull scale parameter	10.0400	2.287e+04	0.60340	16.6300	0.0000
Participant type: PLWH	1.1110	3.037e+00	0.38680	2.8720	0.0041
Sex: male	0.8107	2.250e+00	0.45300	1.7900	0.0735
Age (years)	-0.0563	9.453e-01	0.01918	-2.9350	0.0033
Waist-hip ratio (low)	-0.2816	7.545e-01	0.51790	-0.5438	0.5866
Waist-hip ratio (high)	-0.2046	8.149e-01	0.51270	-0.3991	0.6898

Figure 11.1 shows a Kaplan Meier figure to get a feel of when the events are occurring.

kmSystMod<-survfit(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsSystKM)
censSystModLR<-ictest(Surv(time=survTime1, time2=survTime2, type="interval2") ~ ptype, data = clinEventsSystKM)
logRankPCensSyst<-censSystModLR$p.value

ggplot2:::autoplot(kmSystMod) + 
 labs(x = "\n Time to first clinical event (days)", y = "Probability of no event\n", 
 title = "Kaplan-Meier curves for times until first clinical event by participant type.\nEvents include recorded systemic clinical events as well as new onset diabetes, hypercholesterolemia and chronic kidney conditions.\nThe new onset events are only known up to 6-month interval censoring.",
 caption = paste(sep="","Log-rank test p-value p = ",format(nsmall=4,round(digits=4,logRankPCensSyst)))) +
 theme(plot.title = element_text(hjust = 0.5), 
 axis.title.x = element_text(face="bold", colour="#FF7A33", size = 12),
 axis.title.y = element_text(face="bold", colour="#FF7A33", size = 12),
 legend.title = element_text(face="bold", size = 10)) +
  theme(plot.title.position = "plot",plot.title = element_text(hjust = 0))

Kaplan-Meier curves for times until first systemic clinical event or new-onset diabetes, hypercholesterolemia or chronic kidney condition by participant type.

Figure 11.1: Kaplan-Meier curves for times until first systemic clinical event or new-onset diabetes, hypercholesterolemia or chronic kidney condition by participant type.

12 Mediation analysis

Let \(Y\) be a response of interest, \(X\) a predictor and \(M\) a potential mediator variable. \(M\) is said to be a mediator if:

\(X \rightarrow Y\)
\(X \rightarrow M\)
\(M \rightarrow Y\)

There are direct (not operating through the mediator) and indirect (operating through the mediator) effects of the predictor \(X\) on the outcome \(Y\).

As estimand for a mediation analysis one can focus on controlled direct, natural direct or indirect or stochastic direct or indirect effects (direct and indirect). Each of these will require assumptions: all approaches make consistency and positivity assumptions but differ on assumptions around there being no unmeasured confounding for the various relationships under consideration.

See Rudolph et al. (2018) for more details.

The traditional Baron & Kenny approach to identifying a mediation relationship proceeds as follows:

Regress \(Y\) on \(X\); establish a significant association.
Regress \(M\) on \(X\); establish a significant association.
Regress \(Y\) on both \(X\) and \(M\); establish a significant association between \(Y\) and \(M\) and that the relationship between \(Y\) and \(X\) is now weaker.

There’s further assumptions to be made to be able to interpret this causally and there’s also more sophisticated approaches but they all rely on the same principle: the relationship between \(Y\) and \(X\) is changed if \(M\) is taken into account.

In addition to the 3 variables \(X\), \(Y\) and \(M\), models can be adjusted for other, confounding covariates.

In the case of RHICCA, all of the above is made somewhat more complex by the longitudinal nature of the data. However, the principles of mediation are quite general. What would need to be given care though is whether some effects would be lagged rather than instantaneous.

Here we are interested in exploring the following mediation relationships:

\(Y\) = asymptomatic CVD (continuous CIMT) \(M\) = inflammatory markers (ICAM-1, CD163, IL-6), \(X\) = participant type (HIV)
\(Y\) = asymptomatic CVD (continuous CIMT), \(M\) = herpes virus markers (CMV, VZV), \(X\) = participant type (HIV)
\(Y\) = clinical & sub-clinical CIMT CVD events, \(M\) = inflammatory markers (ICAM-1, CD163, IL-6), \(X\) = participant type (HIV)
\(Y\) = clinical & sub-clinical CIMT CVD events, \(M\) = herpes virus markers (CMV, VZV), \(X\) = participant type (HIV)
Also for systemic events?

The key is to check the 95% confidence interval for the average causal mediation effect (ACME) for each mediation analysis: if this does not include 0, then there is evidence for a mediation.

12.1 Y = asymptomatic CVD (continuous CIMT), M = ICAM-1, X = participant type (HIV)

Our data show an association between X and Y but not X and M and the effect of X on Y does not change when M is included in the model.

This is confirmed by computing and bootstrapping the ACME (and similarly for the natural indirect effect): the 95% confidence interval includes 0, showing no evidence to support a mediation effect.

B<-100
acme<-numeric(B)
estInd<-numeric(B)
tmpLong<-mice::complete(rhiccaDatWide_impMice, action='long', include=TRUE)
nSamp<-unique(table(tmpLong$.imp))
for(b in 1:B){
  idxBoot<-sample(1:nSamp,replace=T,size=nSamp)
  tmpLongBoot<-rbind(tmpLong[tmpLong$.imp==0,][idxBoot,],tmpLong[tmpLong$.imp==1,][idxBoot,])
  for(j in 2:rhiccaDatWide_impMice$m){tmpLongBoot<-rbind(tmpLongBoot,tmpLong[tmpLong$.imp==j,][idxBoot,])}
  rownames(tmpLongBoot)<-NULL
  tmpLongBoot$.id<-rep(1:length(unique(tmpLong$pid)),1+rhiccaDatWide_impMice$m)
  datBoot<-as.mids(tmpLongBoot)
  
  modYX<-summary(pool(with(datBoot,glm(cimtChange~crf.ptype+cimt.cimt_enrolment))))
  modMX<-summary(pool(with(datBoot,glm(lum.ICAM1_change~crf.ptype+lum.ICAM1_enrolment))))
  modYMX<-summary(pool(with(datBoot,glm(cimtChange~crf.ptype+cimt.cimt_enrolment+lum.ICAM1_change))))

  bDirTot<-modYX$estimate[modYX$term=="crf.ptypePatient"]
  bDirMed<-modYMX$estimate[modYMX$term=="crf.ptypePatient"]
  bIndirOut<-modYMX$estimate[modYMX$term=="lum.ICAM1_change"]
  bIndirMed<-modMX$estimate[modMX$term=="crf.ptypePatient"]
  acme[b]<-bDirTot-bDirMed
  estInd[b]<-bIndirOut*bIndirMed
}


modYX<-summary(pool(with(rhiccaDatWide_impMice,glm(cimtChange~crf.ptype+cimt.cimt_enrolment))))
modMX<-summary(pool(with(rhiccaDatWide_impMice,glm(lum.ICAM1_change~crf.ptype+lum.ICAM1_enrolment))))
modYMX<-summary(pool(with(rhiccaDatWide_impMice,glm(cimtChange~crf.ptype+cimt.cimt_enrolment+lum.ICAM1_change))))

bDirTot<-modYX$estimate[modYX$term=="crf.ptypePatient"]
bDirMed<-modYMX$estimate[modYMX$term=="crf.ptypePatient"]

bIndirOut<-modYMX$estimate[modYMX$term=="lum.ICAM1_change"]
bIndirMed<-modMX$estimate[modMX$term=="crf.ptypePatient"]

acmeMainMod<-bDirTot-bDirMed
acmeCI<-quantile(acme,probs=c(0.025,0.975))

indirMainMod<-bIndirOut*bIndirMed
indirCI<-quantile(estInd,probs=c(0.025,0.975))

modOuts<-rbind(modYX,modMX,modYMX)
modOuts %>%
  mutate(
    coefficient=c(c("Intercept","Participant type: PLWH","CIMT (baseline)"),c("Intercept","Participant type: PLWH","ICAM-1 (baseline)"),c("Intercept","Participant type: PLWH","CIMT (baseline)","ICAM-1 (change from baseline)"))
  ) %>%
  dplyr::select(coefficient,estimate,std.error,statistic,df,p.value) %>%
  mutate(p.value=ifelse(coefficient!="Intercept" & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  kableRn(row_names=NULL,row.names=F,escape=F,caption=paste(sep="","Mediation analysis for the effect of participant type through change-from-baseline ICAM-1 on change-from-baseline CIMT. The 95% confidence interval for ACME is given by: (",paste(collapse=", ",format(nsmall=3,round(digits=3,acmeCI))),"). The 95% confidence interval for the natural indirect effect is (",paste(collapse=", ",format(nsmall=3,round(digits=3,indirCI))),").")) %>%
  kable_styling(full_width = FALSE) %>%
  pack_rows(start_row=1,end_row=3,group_label = "model X -> Y") %>%
  pack_rows(start_row=4,end_row=6,group_label = "model X -> M") %>%
  pack_rows(start_row=7,end_row=10,group_label = "model X+M -> Y")

Table 12.1: Mediation analysis for the effect of participant type through change-from-baseline ICAM-1 on change-from-baseline CIMT. The 95% confidence interval for ACME is given by: (0.000, 0.001). The 95% confidence interval for the natural indirect effect is (0.000, 0.001).
coefficient	estimate	std.error	statistic	df	p.value
model X -> Y
Intercept	-9.024600e-03	2.355110e-02	-0.3831938	66.37258	0.7028
Participant type: PLWH	1.635400e-02	6.815900e-03	2.3994041	517.78110	0.0168
CIMT (baseline)	5.118220e-02	4.355430e-02	1.1751343	59.17108	0.2446
model X -> M
Intercept	9.104056e+04	1.455526e+04	6.2548215	70.95405	0.0000
Participant type: PLWH	-2.138729e+04	1.330448e+04	-1.6075254	372.65028	0.1088
ICAM-1 (baseline)	-8.891879e-01	7.042010e-02	-12.6269019	14.83445	0.0000
model X+M -> Y
Intercept	-9.136300e-03	2.353740e-02	-0.3881606	66.88558	0.6991
Participant type: PLWH	1.618370e-02	6.855100e-03	2.3608388	479.80435	0.0186
CIMT (baseline)	5.126270e-02	4.354920e-02	1.1771211	59.38764	0.2438
ICAM-1 (change from baseline)	0.000000e+00	0.000000e+00	-0.4589472	397.33778	0.6465

12.2 Y = asymptomatic CVD (continuous CIMT), M = ICAM-1, CD163, IL-6, X = VZV status (PLWH only)

Our data show an association between X and Y but not X and any of the M variables and the effect of X on Y does not change when the M variables are included in the model.

This is confirmed by computing and bootstrapping the ACME (and similarly for the natural indirect effects): the 95% confidence interval includes 0, showing no evidence to support a mediation effect.

B<-100
acme<-numeric(B)
estInd1<-numeric(B)
estInd2<-numeric(B)
estInd3<-numeric(B)
tmpLong<-mice::complete(rhiccaDatWide_impMice.hiv, action='long', include=TRUE)
nSamp<-unique(table(tmpLong$.imp))
for(b in 1:B){
  idxBoot<-sample(1:nSamp,replace=T,size=nSamp)
  tmpLongBoot<-rbind(tmpLong[tmpLong$.imp==0,][idxBoot,],tmpLong[tmpLong$.imp==1,][idxBoot,])
  for(j in 2:rhiccaDatWide_impMice.hiv$m){tmpLongBoot<-rbind(tmpLongBoot,tmpLong[tmpLong$.imp==j,][idxBoot,])}
  rownames(tmpLongBoot)<-NULL
  tmpLongBoot$.id<-rep(1:length(unique(tmpLong$pid)),1+rhiccaDatWide_impMice.hiv$m)
  datBoot<-as.mids(tmpLongBoot)
  
  modYX<-summary(pool(with(datBoot,glm(cimtChange~vzvStatus+cimt.cimt_enrolment))))
  modM1X<-summary(pool(with(datBoot,glm(log(lum.ICAM1_enrolment)~vzvStatus))))
  modM2X<-summary(pool(with(datBoot,glm(log(lum.CD163_enrolment)~vzvStatus))))
  modM3X<-summary(pool(with(datBoot,glm(log(lum.IL6_enrolment)~vzvStatus))))
  modYMX<-summary(pool(with(datBoot,glm(cimtChange~vzvStatus+cimt.cimt_enrolment+log(lum.ICAM1_enrolment)+log(lum.CD163_enrolment)+log(lum.IL6_enrolment)))))

  bDirTot<-modYX$estimate[modYX$term=="vzvStatuspersistently high titres"]
  bDirMed<-modYMX$estimate[modYMX$term=="vzvStatuspersistently high titres"]
  bIndir1Out<-modYMX$estimate[modYMX$term=="log(lum.ICAM1_enrolment)"]
  bIndir1Med<-modM1X$estimate[modM1X$term=="vzvStatuspersistently high titres"]
  bIndir2Out<-modYMX$estimate[modYMX$term=="log(lum.CD163_enrolment)"]
  bIndir2Med<-modM2X$estimate[modM2X$term=="vzvStatuspersistently high titres"]
  bIndir3Out<-modYMX$estimate[modYMX$term=="log(lum.IL6_enrolment)"]
  bIndir3Med<-modM3X$estimate[modM3X$term=="vzvStatuspersistently high titres"]
  acme[b]<-bDirTot-bDirMed
  estInd1[b]<-bIndir1Out*bIndir1Med
  estInd2[b]<-bIndir2Out*bIndir2Med
  estInd3[b]<-bIndir3Out*bIndir3Med
}

modYX<-summary(pool(with(rhiccaDatWide_impMice.hiv,glm(cimtChange~vzvStatus+cimt.cimt_enrolment))))
modM1X<-summary(pool(with(rhiccaDatWide_impMice.hiv,glm(log(lum.ICAM1_enrolment)~vzvStatus))))
modM2X<-summary(pool(with(rhiccaDatWide_impMice.hiv,glm(log(lum.CD163_enrolment)~vzvStatus))))
modM3X<-summary(pool(with(rhiccaDatWide_impMice.hiv,glm(log(lum.IL6_enrolment)~vzvStatus))))
modYMX<-summary(pool(with(rhiccaDatWide_impMice.hiv,glm(cimtChange~vzvStatus+cimt.cimt_enrolment+log(lum.ICAM1_enrolment)+log(lum.CD163_enrolment)+log(lum.IL6_enrolment)))))

bDirTot<-modYX$estimate[modYX$term=="vzvStatuspersistently high titres"]
bDirMed<-modYMX$estimate[modYMX$term=="vzvStatuspersistently high titres"]
bIndir1Out<-modYMX$estimate[modYMX$term=="log(lum.ICAM1_enrolment)"]
bIndir1Med<-modM1X$estimate[modM1X$term=="vzvStatuspersistently high titres"]
bIndir2Out<-modYMX$estimate[modYMX$term=="log(lum.CD163_enrolment)"]
bIndir2Med<-modM2X$estimate[modM2X$term=="vzvStatuspersistently high titres"]
bIndir3Out<-modYMX$estimate[modYMX$term=="log(lum.IL6_enrolment)"]
bIndir3Med<-modM3X$estimate[modM3X$term=="vzvStatuspersistently high titres"]

acmeMainMod<-bDirTot-bDirMed
acmeCI<-quantile(acme,probs=c(0.025,0.975))

indir1MainMod<-bIndir1Out*bIndir1Med
indir2MainMod<-bIndir2Out*bIndir2Med
indir3MainMod<-bIndir3Out*bIndir3Med
indir1CI<-quantile(estInd1,probs=c(0.025,0.975))
indir2CI<-quantile(estInd2,probs=c(0.025,0.975))
indir3CI<-quantile(estInd3,probs=c(0.025,0.975))

modOuts<-rbind(modYX,modM1X,modM2X,modM3X,modYMX)
modOuts %>%
  mutate(
    coefficient=c(c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres","CIMT (baseline)"),c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres","CIMT (baseline)","log ICAM-1 (baseline)","log CD163 (baseline)","log IL-6 (baseline)"))
  ) %>%
  dplyr::select(coefficient,estimate,std.error,statistic,df,p.value) %>%
  mutate(p.value=ifelse(coefficient!="Intercept" & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  kableRn(row_names=NULL,row.names=F,escape=F,caption=paste(sep="","Mediation analysis for the effect of VZV status through baseline inflammatory markers on change-from-baseline CIMT (PLWH only). The 95% confidence interval for ACME is given by: (",paste(collapse=", ",format(nsmall=3,round(digits=3,acmeCI))),"). The 95% CI for the 3 natural indirect effects of ICAM-1, CD163 and IL-6 are given by (",paste(collapse=", ",format(nsmall=3,round(digits=3,indir1CI))),"), (",paste(collapse=", ",format(nsmall=3,round(digits=3,indir2CI))),") and (",paste(collapse=", ",format(nsmall=3,round(digits=3,indir3CI))),").")) %>%
  kable_styling(full_width = FALSE) %>%
  pack_rows(start_row=1,end_row=5,group_label = "model X -> Y") %>%
  pack_rows(start_row=6,end_row=9,group_label = "model X -> M (ICAM-1)") %>%
  pack_rows(start_row=10,end_row=13,group_label = "model X -> M (CD163)") %>%
  pack_rows(start_row=14,end_row=17,group_label = "model X -> M (IL-6)") %>%
  pack_rows(start_row=18,end_row=25,group_label = "model X+M -> Y")

Table 12.2: Mediation analysis for the effect of VZV status through baseline inflammatory markers on change-from-baseline CIMT (PLWH only). The 95% confidence interval for ACME is given by: (-0.001, 0.002). The 95% CI for the 3 natural indirect effects of ICAM-1, CD163 and IL-6 are given by (-0.001, 0.001), (-0.002, 0.001) and (0.000, 0.001).
coefficient	estimate	std.error	statistic	df	p.value
model X -> Y
Intercept	-0.0221326	0.0267480	-0.8274500	91.30846	0.4101
VZV: progression to latency	0.0221606	0.0116774	1.8977320	97.98794	0.0607
VZV: reactivation	0.0039354	0.0155883	0.2524579	309.93213	0.8009
VZV: persistently high titres	0.0197700	0.0091696	2.1560401	289.06696	0.0319
CIMT (baseline)	0.0966808	0.0541096	1.7867582	69.43050	0.0783
model X -> M (ICAM-1)
Intercept	11.1967995	0.0520230	215.2278245	413.75236	0.0000
VZV: progression to latency	-0.0345259	0.1362411	-0.2534176	206.89368	0.8002
VZV: reactivation	0.0441548	0.2257248	0.1956135	65.68688	0.8455
VZV: persistently high titres	0.0533616	0.1194097	0.4468784	156.94341	0.6556
model X -> M (CD163)
Intercept	4.4709358	0.0462295	96.7117535	354.79926	0.0000
VZV: progression to latency	0.0466885	0.1248151	0.3740614	126.72971	0.7090
VZV: reactivation	-0.1279141	0.2018560	-0.6336896	58.95432	0.5287
VZV: persistently high titres	0.0497791	0.1223808	0.4067557	47.07838	0.6860
model X -> M (IL-6)
Intercept	0.8163182	0.0770376	10.5963555	67.35016	0.0000
VZV: progression to latency	0.1333728	0.1923942	0.6932266	68.02669	0.4905
VZV: reactivation	0.3610922	0.2645358	1.3650030	121.23102	0.1748
VZV: persistently high titres	0.1503503	0.1634218	0.9200132	73.03403	0.3606
model X+M -> Y
Intercept	-0.0311183	0.0445378	-0.6986942	242.13010	0.4854
VZV: progression to latency	0.0220470	0.0116913	1.8857573	98.10594	0.0623
VZV: reactivation	0.0022971	0.0156642	0.1466455	303.89534	0.8835
VZV: persistently high titres	0.0194567	0.0091606	2.1239597	302.26393	0.0345
CIMT (baseline)	0.0992796	0.0539547	1.8400555	71.94369	0.0699
log ICAM-1 (baseline)	0.0020192	0.0039793	0.5074215	90.02038	0.6131
log CD163 (baseline)	-0.0038604	0.0040030	-0.9643800	474.20978	0.3353
log IL-6 (baseline)	0.0028465	0.0027225	1.0455485	155.29901	0.2974

tmp<-modOuts %>%
  mutate(
    coefficient=c(c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres","CIMT (baseline)"),c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres","CIMT (baseline)","log ICAM-1 (baseline)","log CD163 (baseline)","log IL-6 (baseline)"))
  ) %>%
  dplyr::select(coefficient,estimate,std.error,statistic,df,p.value) 

write.csv(tmp,file=paste(sep="_",outPrefix,"Mediation_YcimtMinflammatoryXvzv_Table4a.csv"))

12.3 Y = clinical and subclinical (CIMT) CVD events, M = ICAM-1, CD163, IL-6, X = VZV status (PLWH only)

Our data show an association between X and Y but not X and any of the M variables and the effect of X on Y does not change when the M variables are included in the model.

This is confirmed by computing and bootstrapping the ACME (and similarly for the natural indirect effects): the 95% confidence interval includes 0, showing no evidence to support a mediation effect.

B<-100
acme<-numeric(B)
estInd1<-numeric(B)
estInd2<-numeric(B)
estInd3<-numeric(B)
tmpLong<-mice::complete(rhiccaDatWide_impMice.hiv, action='long', include=TRUE)
tmpLong2<-mice::complete(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"), action='long', include=TRUE)
nSamp<-unique(table(tmpLong$.imp))
m<-rhiccaDatWide_impMice.hiv$m
nSamp2<-unique(table(tmpLong2$.imp))
m2<-rhiccaDatWide_impMice.clinEventsConfirmed$m
if(nSamp!=nSamp2){stop("nSamp and nSamp2 should be the same!")}
if(m!=m2){stop("m and m2 should be the same!")}
for(b in 1:B){
  idxBoot<-sample(1:nSamp,replace=T,size=nSamp)
  tmpLongBoot<-rbind(tmpLong[tmpLong$.imp==0,][idxBoot,],tmpLong[tmpLong$.imp==1,][idxBoot,])
  tmpLong2Boot<-rbind(tmpLong2[tmpLong2$.imp==0,][idxBoot,],tmpLong2[tmpLong2$.imp==1,][idxBoot,])
  for(j in 2:m){
    tmpLongBoot<-rbind(tmpLongBoot,tmpLong[tmpLong$.imp==j,][idxBoot,])
    tmpLong2Boot<-rbind(tmpLong2Boot,tmpLong2[tmpLong2$.imp==j,][idxBoot,])
  }
  rownames(tmpLongBoot)<-NULL
  rownames(tmpLong2Boot)<-NULL
  tmpLongBoot$.id<-rep(1:length(unique(tmpLong$pid)),1+m)
  tmpLong2Boot$.id<-rep(1:length(unique(tmpLong2$pid)),1+m)
  datBoot<-as.mids(tmpLongBoot)
  datBoot2<-as.mids(tmpLong2Boot)
  
  modYX<-summary(pool(with(datBoot2,glm(numClinEventsPerParticipantCVDwithCIMT~crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+vzvStatus,offset=obsTime/(1000*365.25),family="quasipoisson"))))
  modM1X<-summary(pool(with(datBoot,glm(log(lum.ICAM1_enrolment)~vzvStatus))))
  modM2X<-summary(pool(with(datBoot,glm(log(lum.CD163_enrolment)~vzvStatus))))
  modM3X<-summary(pool(with(datBoot,glm(log(lum.IL6_enrolment)~vzvStatus))))
  modYMX<-summary(pool(with(datBoot2,glm(numClinEventsPerParticipantCVDwithCIMT~crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+vzvStatus+log(lum.ICAM1_enrolment)+log(lum.CD163_enrolment)+log(lum.IL6_enrolment),offset=obsTime/(1000*365.25),family="quasipoisson"))))

  bDirTot<-modYX$estimate[modYX$term=="vzvStatuspersistently high titres"]
  bDirMed<-modYMX$estimate[modYMX$term=="vzvStatuspersistently high titres"]
  bIndir1Out<-modYMX$estimate[modYMX$term=="log(lum.ICAM1_enrolment)"]
  bIndir1Med<-modM1X$estimate[modM1X$term=="vzvStatuspersistently high titres"]
  bIndir2Out<-modYMX$estimate[modYMX$term=="log(lum.CD163_enrolment)"]
  bIndir2Med<-modM2X$estimate[modM2X$term=="vzvStatuspersistently high titres"]
  bIndir3Out<-modYMX$estimate[modYMX$term=="log(lum.IL6_enrolment)"]
  bIndir3Med<-modM3X$estimate[modM3X$term=="vzvStatuspersistently high titres"]
  acme[b]<-bDirTot-bDirMed
  estInd1[b]<-bIndir1Out*bIndir1Med
  estInd2[b]<-bIndir2Out*bIndir2Med
  estInd3[b]<-bIndir3Out*bIndir3Med
}

modYX<-summary(pool(with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(numClinEventsPerParticipantCVDwithCIMT~crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+vzvStatus,offset=obsTime/(1000*365.25),family="quasipoisson"))))
modM1X<-summary(pool(with(rhiccaDatWide_impMice.hiv,glm(log(lum.ICAM1_enrolment)~vzvStatus))))
modM2X<-summary(pool(with(rhiccaDatWide_impMice.hiv,glm(log(lum.CD163_enrolment)~vzvStatus))))
modM3X<-summary(pool(with(rhiccaDatWide_impMice.hiv,glm(log(lum.IL6_enrolment)~vzvStatus))))
modYMX<-summary(pool(with(rhiccaDatWide_impMice.clinEventsConfirmed %>% dplyr::filter(crf.ptype=="Patient"),glm(numClinEventsPerParticipantCVDwithCIMT~crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+vzvStatus+log(lum.ICAM1_enrolment)+log(lum.CD163_enrolment)+log(lum.IL6_enrolment),offset=obsTime/(1000*365.25),family="quasipoisson"))))

bDirTot<-modYX$estimate[modYX$term=="vzvStatuspersistently high titres"]
bDirMed<-modYMX$estimate[modYMX$term=="vzvStatuspersistently high titres"]
bIndir1Out<-modYMX$estimate[modYMX$term=="log(lum.ICAM1_enrolment)"]
bIndir1Med<-modM1X$estimate[modM1X$term=="vzvStatuspersistently high titres"]
bIndir2Out<-modYMX$estimate[modYMX$term=="log(lum.CD163_enrolment)"]
bIndir2Med<-modM2X$estimate[modM2X$term=="vzvStatuspersistently high titres"]
bIndir3Out<-modYMX$estimate[modYMX$term=="log(lum.IL6_enrolment)"]
bIndir3Med<-modM3X$estimate[modM3X$term=="vzvStatuspersistently high titres"]

acmeMainMod<-bDirTot-bDirMed
acmeCI<-quantile(acme,probs=c(0.025,0.975))

indir1MainMod<-bIndir1Out*bIndir1Med
indir2MainMod<-bIndir2Out*bIndir2Med
indir3MainMod<-bIndir3Out*bIndir3Med
indir1CI<-quantile(estInd1,probs=c(0.025,0.975))
indir2CI<-quantile(estInd2,probs=c(0.025,0.975))
indir3CI<-quantile(estInd3,probs=c(0.025,0.975))

modOuts<-rbind(modYX,modM1X,modM2X,modM3X,modYMX)
modOuts %>%
  mutate(
    coefficient=c(c("Intercept","Sex male","Age","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),rep(c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),3),c("Intercept","Sex male","Age","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres","log ICAM-1 (baseline)","log CD163 (baseline)","log IL-6 (baseline)"))
  ) %>%
  dplyr::select(coefficient,estimate,std.error,statistic,df,p.value) %>%
  mutate(p.value=ifelse(coefficient!="Intercept" & p.value<0.05,cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=T),cell_spec(format(nsmall=4,round(digits=4,p.value)),bold=F))) %>%
  kableRn(row_names=NULL,row.names=F,escape=F,caption=paste(sep="","Mediation analysis for the effect of VZV status through baseline inflammatory markers on clinical and sub-clinical (CIMT) CVD events. The 95% confidence interval for ACME is given by: (",paste(collapse=", ",format(nsmall=3,round(digits=3,acmeCI))),"). The 95% CI for the 3 natural indirect effects of ICAM-1, CD163 and IL-6 are given by (",paste(collapse=", ",format(nsmall=3,round(digits=3,indir1CI))),"), (",paste(collapse=", ",format(nsmall=3,round(digits=3,indir2CI))),") and (",paste(collapse=", ",format(nsmall=3,round(digits=3,indir3CI))),").")) %>%
  kable_styling(full_width = FALSE) %>%
  pack_rows(start_row=1,end_row=11,group_label = "model X -> Y") %>%
  pack_rows(start_row=12,end_row=15,group_label = "model X -> M (ICAM-1)") %>%
  pack_rows(start_row=16,end_row=19,group_label = "model X -> M (CD163)") %>%
  pack_rows(start_row=20,end_row=23,group_label = "model X -> M (IL-6)") %>%
  pack_rows(start_row=24,end_row=37,group_label = "model X+M -> Y")

Table 12.3: Mediation analysis for the effect of VZV status through baseline inflammatory markers on clinical and sub-clinical (CIMT) CVD events. The 95% confidence interval for ACME is given by: (-0.210, 0.435). The 95% CI for the 3 natural indirect effects of ICAM-1, CD163 and IL-6 are given by (-0.055, 0.102), (-0.045, 0.074) and (-0.063, 0.042).
coefficient	estimate	std.error	statistic	df	p.value
model X -> Y
Intercept	-10.8361176	2.2942172	-4.7232309	652.33354	0.0000
Sex male	-0.4261672	0.5070599	-0.8404671	690.22605	0.4009
Age	0.1157578	0.0233150	4.9649552	654.44567	0.0000
BMI	0.0540901	0.0538911	1.0036927	669.77182	0.3159
Diastolic blood pressure	0.0278238	0.0361471	0.7697379	692.17931	0.4417
Systolic blood pressure	-0.0044602	0.0202929	-0.2197901	692.04773	0.8261
Glucose (baseline)	-0.0804439	0.1671041	-0.4814003	684.83751	0.6304
Cholesterol (baseline)	-0.2849331	0.2129771	-1.3378578	685.05892	0.1814
VZV: progression to latency	-0.2869483	0.9519995	-0.3014164	210.55641	0.7634
VZV: reactivation	-6.6334879	865.4982673	-0.0076644	696.93804	0.9939
VZV: persistently high titres	1.1685419	0.4985307	2.3439719	323.39877	0.0197
model X -> M (ICAM-1)
Intercept	11.1967995	0.0520230	215.2278245	413.75236	0.0000
VZV: progression to latency	-0.0345259	0.1362411	-0.2534176	206.89368	0.8002
VZV: reactivation	0.0441548	0.2257248	0.1956135	65.68688	0.8455
VZV: persistently high titres	0.0533616	0.1194097	0.4468784	156.94341	0.6556
model X -> M (CD163)
Intercept	4.4709358	0.0462295	96.7117535	354.79926	0.0000
VZV: progression to latency	0.0466885	0.1248151	0.3740614	126.72971	0.7090
VZV: reactivation	-0.1279141	0.2018560	-0.6336896	58.95432	0.5287
VZV: persistently high titres	0.0497791	0.1223808	0.4067557	47.07838	0.6860
model X -> M (IL-6)
Intercept	0.8163182	0.0770376	10.5963555	67.35016	0.0000
VZV: progression to latency	0.1333728	0.1923942	0.6932266	68.02669	0.4905
VZV: reactivation	0.3610922	0.2645358	1.3650030	121.23102	0.1748
VZV: persistently high titres	0.1503503	0.1634218	0.9200132	73.03403	0.3606
model X+M -> Y
Intercept	-13.6067377	3.4447663	-3.9499741	380.06411	0.0001
Sex male	-0.4136555	0.4995888	-0.8279919	681.22151	0.4080
Age	0.1187297	0.0243020	4.8855852	609.88485	0.0000
BMI	0.0554191	0.0542189	1.0221360	632.66430	0.3071
Diastolic blood pressure	0.0143531	0.0376271	0.3814568	688.65581	0.7030
Systolic blood pressure	0.0040485	0.0216251	0.1872148	677.99552	0.8515
Glucose (baseline)	-0.0658565	0.1506520	-0.4371436	658.27419	0.6622
Cholesterol (baseline)	-0.2348602	0.2273020	-1.0332521	621.13520	0.3019
VZV: progression to latency	-0.4535073	0.9580223	-0.4733786	197.77731	0.6365
VZV: reactivation	-6.7528177	849.9980090	-0.0079445	693.93714	0.9937
VZV: persistently high titres	0.9797346	0.5145522	1.9040527	334.33662	0.0578
log ICAM-1 (baseline)	0.1319484	0.2406412	0.5483201	176.76827	0.5842
log CD163 (baseline)	0.2075722	0.2317415	0.8957059	320.28122	0.3711
log IL-6 (baseline)	-0.0133401	0.1976327	-0.0674997	87.70231	0.9463

tmp<-modOuts %>%
  mutate(
    coefficient=c(c("Intercept","Sex male","Age","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),rep(c("Intercept","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres"),3),c("Intercept","Sex male","Age","BMI","Diastolic blood pressure","Systolic blood pressure","Glucose (baseline)","Cholesterol (baseline)","VZV: progression to latency","VZV: reactivation","VZV: persistently high titres","log ICAM-1 (baseline)","log CD163 (baseline)","log IL-6 (baseline)"))
  ) %>%
  dplyr::select(coefficient,estimate,std.error,statistic,df,p.value)

write.csv(tmp,file=paste(sep="_",outPrefix,"Mediation_YeventsMinflammatoryXvzv_Table4b.csv"))

13 Additional things for discussion

rhiccaDatWide <- rhiccaDatWide %>%
  dplyr::mutate(
    cimt.cimt_changeFromBaseline=cimt.cimt_24months-cimt.cimt_enrolment,
    cimt.cimt_changeFromBaselineRelative=(cimt.cimt_24months-cimt.cimt_enrolment)/cimt.cimt_enrolment
  )

cimtRelChangeSummary <- rhiccaDatWide %>%
  dplyr::filter(!is.na(crf.PLWH) & !is.na(vzvStatus_2levels)) %>%
  dplyr::group_by(crf.PLWH,vzvStatus_2levels) %>%
  dplyr::summarise(
    minCimtRelChange=paste(sep="",format(nsmall=1,round(digits=1,100*min(cimt.cimt_changeFromBaselineRelative,na.rm=TRUE))),"%"),
    maxCimtRelChange=paste(sep="",format(nsmall=1,round(digits=1,100*max(cimt.cimt_changeFromBaselineRelative,na.rm=TRUE))),"%"),
    median=paste(sep="",format(nsmall=1,round(digits=1,100*quantile(probs=0.5,cimt.cimt_changeFromBaselineRelative,na.rm=TRUE))),"%"),
    q25CimtRelChange=paste(sep="",format(nsmall=1,round(digits=1,100*quantile(probs=0.25,cimt.cimt_changeFromBaselineRelative,na.rm=TRUE))),"%"),
    q75CimtRelChange=paste(sep="",format(nsmall=1,round(digits=1,100*quantile(probs=0.75,cimt.cimt_changeFromBaselineRelative,na.rm=TRUE))),"%")
  )

cimtRelChangeSummary %>%
  kable(row.names=F,col.names=c("Participant type","VZV status","Minimum","Maximum","Median","25th percentile","75th percentile")) %>%
  kable_styling(full_width = FALSE) %>%
  add_header_above(header=c(" "=2,"Relative change in CIMT from enrolment to 24 months"=5))

		Relative change in CIMT from enrolment to 24 months
Participant type	VZV status	Minimum	Maximum	Median	25th percentile	75th percentile
Control	not persistently high titres	-10.1%	23.3%	3.5%	-1.3%	8.9%
Control	persistently high titres	-14.9%	13.9%	0.5%	-1.6%	2.6%
PLWH	not persistently high titres	-15.7%	198.7%	4.7%	0.4%	10.3%
PLWH	persistently high titres	-17.6%	206.1%	7.1%	4.1%	11.9%

tmp<-complete(rhiccaDatWide_impMice,action = "long",include=TRUE)
tmp<-tmp %>% mutate(cimtChange=cimt.cimt_24months-cimt.cimt_enrolment)
tmp.mids<-as.mids(tmp)

tt<-with(tmp.mids %>% filter(crf.ptype=="Patient"),lm(cimtChange~1))
ttChange<-summary(pool(tt))$estimate
tt<-with(tmp.mids %>% filter(crf.ptype=="Patient"),lm(cimt.cimt_enrolment~1))
ttBaseline<-summary(pool(tt))$estimate
cat(paste(sep="","Proportion change in CIMT from baseline to 24 months in PLWH: ",ttChange/ttBaseline,"\n"))

## Proportion change in CIMT from baseline to 24 months in PLWH: 0.0656813550256049

tt<-with(tmp.mids %>% filter(crf.ptype=="Control"),lm(cimtChange~1))
ttChange<-summary(pool(tt))$estimate
tt<-with(tmp.mids %>% filter(crf.ptype=="Control"),lm(cimt.cimt_enrolment~1))
ttBaseline<-summary(pool(tt))$estimate
cat(paste(sep="","Proportion change in CIMT from baseline to 24 months in Controls: ",ttChange/ttBaseline,"\n"))

## Proportion change in CIMT from baseline to 24 months in Controls: 0.0339636189564385

tt<-with(tmp.mids,lm(cimtChange~crf.ptype)); summary(pool(tt))

##               term   estimate   std.error statistic       df     p.value
## 1      (Intercept) 0.01780108 0.005892811  3.020812 784.5559 0.002602819
## 2 crf.ptypePatient 0.01540503 0.006711295  2.295388 636.4445 0.022035231

fisher.test(table(rhiccaDatWide$crf.artRegimen.2levels,rhiccaDatWide$cmvStatus_2levels))

## 
##  Fisher's Exact Test for Count Data
## 
## data:  table(rhiccaDatWide$crf.artRegimen.2levels, rhiccaDatWide$cmvStatus_2levels)
## p-value = 0.006234
## alternative hypothesis: true odds ratio is not equal to 1
## 95 percent confidence interval:
##  1.192250 3.240689
## sample estimates:
## odds ratio 
##   1.957222

fisher.test(table(rhiccaDatWide$crf.artRegimen.2levels,rhiccaDatWide$vzvStatus_2levels))

## 
##  Fisher's Exact Test for Count Data
## 
## data:  table(rhiccaDatWide$crf.artRegimen.2levels, rhiccaDatWide$vzvStatus_2levels)
## p-value = 0.4416
## alternative hypothesis: true odds ratio is not equal to 1
## 95 percent confidence interval:
##  0.7248986 2.1164024
## sample estimates:
## odds ratio 
##   1.237598

sumFitandSum_pwv.hiv("visitNum+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia+physio.bp_sys+lims.Glucose_enrolment+lims.Cholesterol_enrolment+crf.artRegimen.2levels*cmvStatus+crf.artRegimen.2levels*vzvStatus+lims.CD4_Absolute_enrolment+lims.CD4_Absolute")

##                                                                                        term
## 1                                                                               (Intercept)
## 2                                                                                  visitNum
## 3                                                                            crf.genderMale
## 4                                                                                   crf.age
## 5                                                                      physio.bmi_enrolment
## 6                                                                             physio.bp_dia
## 7                                                                             physio.bp_sys
## 8                                                                    lims.Glucose_enrolment
## 9                                                                lims.Cholesterol_enrolment
## 10                                     crf.artRegimen.2levelsART: No change to dolutegravir
## 11                                                       cmvStatusprogression to low titres
## 12                                                      cmvStatusprogression to high titres
## 13                                                        cmvStatuspersistently high titres
## 14                                                       vzvStatusprogression to low titres
## 15                                                      vzvStatusprogression to high titres
## 16                                                        vzvStatuspersistently high titres
## 17                                                              lims.CD4_Absolute_enrolment
## 18                                                                        lims.CD4_Absolute
## 19  crf.artRegimen.2levelsART: No change to dolutegravir:cmvStatusprogression to low titres
## 20 crf.artRegimen.2levelsART: No change to dolutegravir:cmvStatusprogression to high titres
## 21   crf.artRegimen.2levelsART: No change to dolutegravir:cmvStatuspersistently high titres
## 22  crf.artRegimen.2levelsART: No change to dolutegravir:vzvStatusprogression to low titres
## 23 crf.artRegimen.2levelsART: No change to dolutegravir:vzvStatusprogression to high titres
## 24   crf.artRegimen.2levelsART: No change to dolutegravir:vzvStatuspersistently high titres
##         estimate    std.error    statistic         df
## 1   2.6960933219 0.4964329904  5.430931010  167.45512
## 2  -0.0139897083 0.0040250602 -3.475651912  174.87268
## 3  -0.0655520336 0.1096009825 -0.598097135  202.21080
## 4   0.0585243481 0.0070590719  8.290657637   99.78442
## 5  -0.0129748745 0.0116066203 -1.117885668  890.71522
## 6   0.0325070543 0.0062791295  5.177000179  206.41366
## 7   0.0105628624 0.0041254311  2.560426356  295.46666
## 8   0.0504509257 0.0230177153  2.191830299 1458.82451
## 9  -0.0002862432 0.0485945127 -0.005890443  234.38455
## 10 -0.1063067592 0.1352326905 -0.786102523  541.26387
## 11 -0.2413023963 0.4101445803 -0.588334963  296.81750
## 12  0.0399282171 0.2033544134  0.196347925  189.09839
## 13 -0.1285914668 0.1794056863 -0.716763607  528.24386
## 14  0.1623440984 0.2138026935  0.759317368  125.08593
## 15 -0.1159693883 0.2794586727 -0.414978670  369.33895
## 16  0.1274865394 0.1856355974  0.686756965  273.61632
## 17 -0.0002371072 0.0002886273 -0.821499582  299.36451
## 18 -0.0008768649 0.0002744481 -3.195012283  226.24239
## 19  0.0216429202 0.6992247708  0.030952737  153.45071
## 20  0.1062921741 0.2867134008  0.370726216  104.14421
## 21  0.3194278202 0.2324062794  1.374437133  508.98497
## 22 -0.0593641150 0.3005143386 -0.197541706  199.32572
## 23  0.3205949627 0.4561277128  0.702862277  165.28714
## 24 -0.2364262793 0.2485120773 -0.951367361  366.04998
##                                                 p.value rownum
## 1                    <span style="     " >0.0000</span>      1
## 2  <span style=" font-weight: bold;    " >0.0006</span>      2
## 3                    <span style="     " >0.5504</span>      3
## 4  <span style=" font-weight: bold;    " >0.0000</span>      4
## 5                    <span style="     " >0.2639</span>      5
## 6  <span style=" font-weight: bold;    " >0.0000</span>      6
## 7  <span style=" font-weight: bold;    " >0.0110</span>      7
## 8  <span style=" font-weight: bold;    " >0.0285</span>      8
## 9                    <span style="     " >0.9953</span>      9
## 10                   <span style="     " >0.4322</span>     10
## 11                   <span style="     " >0.5568</span>     11
## 12                   <span style="     " >0.8445</span>     12
## 13                   <span style="     " >0.4738</span>     13
## 14                   <span style="     " >0.4491</span>     14
## 15                   <span style="     " >0.6784</span>     15
## 16                   <span style="     " >0.4928</span>     16
## 17                   <span style="     " >0.4120</span>     17
## 18 <span style=" font-weight: bold;    " >0.0016</span>     18
## 19                   <span style="     " >0.9753</span>     19
## 20                   <span style="     " >0.7116</span>     20
## 21                   <span style="     " >0.1699</span>     21
## 22                   <span style="     " >0.8436</span>     22
## 23                   <span style="     " >0.4831</span>     23
## 24                   <span style="     " >0.3420</span>     24

sumFitandSum_cimt.hiv("+crf.gender+crf.age+physio.bmi_enrolment+physio.bp_dia_enrolment+physio.bp_sys_enrolment+lims.Glucose_enrolment+lims.Cholesterol_enrolment+crf.artRegimen.2levels*cmvStatus+crf.artRegimen.2levels*vzvStatus+lims.CD4_Absolute_enrolment+lims.CD4_Absolute_changeFromBaselineAt24Months")

##                                                                                        term
## 1                                                                               (Intercept)
## 2                                                                       cimt.cimt_enrolment
## 3                                                                            crf.genderMale
## 4                                                                                   crf.age
## 5                                                                      physio.bmi_enrolment
## 6                                                                   physio.bp_dia_enrolment
## 7                                                                   physio.bp_sys_enrolment
## 8                                                                    lims.Glucose_enrolment
## 9                                                                lims.Cholesterol_enrolment
## 10                                     crf.artRegimen.2levelsART: No change to dolutegravir
## 11                                                       cmvStatusprogression to low titres
## 12                                                      cmvStatusprogression to high titres
## 13                                                        cmvStatuspersistently high titres
## 14                                                       vzvStatusprogression to low titres
## 15                                                      vzvStatusprogression to high titres
## 16                                                        vzvStatuspersistently high titres
## 17                                                              lims.CD4_Absolute_enrolment
## 18                                           lims.CD4_Absolute_changeFromBaselineAt24Months
## 19  crf.artRegimen.2levelsART: No change to dolutegravir:cmvStatusprogression to low titres
## 20 crf.artRegimen.2levelsART: No change to dolutegravir:cmvStatusprogression to high titres
## 21   crf.artRegimen.2levelsART: No change to dolutegravir:cmvStatuspersistently high titres
## 22  crf.artRegimen.2levelsART: No change to dolutegravir:vzvStatusprogression to low titres
## 23 crf.artRegimen.2levelsART: No change to dolutegravir:vzvStatusprogression to high titres
## 24   crf.artRegimen.2levelsART: No change to dolutegravir:vzvStatuspersistently high titres
##         estimate    std.error   statistic        df
## 1  -2.448768e-02 3.833653e-02 -0.63875579 167.51374
## 2   5.359526e-02 7.101882e-02  0.75466283  50.94333
## 3  -5.432029e-03 7.995598e-03 -0.67937750 338.65251
## 4   9.087242e-04 7.585171e-04  1.19802727  29.64629
## 5  -4.865285e-04 9.485379e-04 -0.51292474 210.24100
## 6  -4.293818e-04 6.125645e-04 -0.70095774 348.38838
## 7   3.112047e-04 3.964487e-04  0.78498098 417.72815
## 8   7.503244e-04 1.769935e-03  0.42392774 401.80314
## 9  -5.334801e-03 3.721597e-03 -1.43347062 151.02365
## 10 -1.152859e-03 1.017093e-02 -0.11334843 310.58630
## 11 -8.524916e-04 2.972865e-02 -0.02867576 384.88066
## 12  3.001874e-03 1.404830e-02  0.21368239 489.02827
## 13  2.796022e-02 1.494339e-02  1.87107564  90.52844
## 14  6.881450e-03 1.583377e-02  0.43460597 115.21406
## 15  7.339145e-03 2.081300e-02  0.35262315 263.92234
## 16  3.319991e-02 1.384744e-02  2.39754894 212.78774
## 17  1.363597e-05 2.478134e-05  0.55025166  69.78186
## 18  2.862965e-05 2.438828e-05  1.17390995  84.06231
## 19 -6.189654e-03 4.716040e-02 -0.13124685 530.23410
## 20  2.224912e-02 1.965128e-02  1.13219691 242.45274
## 21 -3.186829e-02 1.944354e-02 -1.63901690  88.99383
## 22  2.900351e-02 2.367214e-02  1.22521698  91.35905
## 23 -6.428736e-03 3.236677e-02 -0.19862147 251.63579
## 24 -2.470207e-02 1.895602e-02 -1.30312521 194.20717
##                                                p.value rownum
## 1                    <span style="     " >0.524</span>      1
## 2                    <span style="     " >0.454</span>      2
## 3                    <span style="     " >0.497</span>      3
## 4                    <span style="     " >0.240</span>      4
## 5                    <span style="     " >0.609</span>      5
## 6                    <span style="     " >0.484</span>      6
## 7                    <span style="     " >0.433</span>      7
## 8                    <span style="     " >0.672</span>      8
## 9                    <span style="     " >0.154</span>      9
## 10                   <span style="     " >0.910</span>     10
## 11                   <span style="     " >0.977</span>     11
## 12                   <span style="     " >0.831</span>     12
## 13                   <span style="     " >0.065</span>     13
## 14                   <span style="     " >0.665</span>     14
## 15                   <span style="     " >0.725</span>     15
## 16 <span style=" font-weight: bold;    " >0.017</span>     16
## 17                   <span style="     " >0.584</span>     17
## 18                   <span style="     " >0.244</span>     18
## 19                   <span style="     " >0.896</span>     19
## 20                   <span style="     " >0.259</span>     20
## 21                   <span style="     " >0.105</span>     21
## 22                   <span style="     " >0.224</span>     22
## 23                   <span style="     " >0.843</span>     23
## 24                   <span style="     " >0.194</span>     24

# identify those who are virally suppressed and not-suppressed
rhiccaDatWide<-rhiccaDatWide %>%
  dplyr::mutate(
    cimtChange=cimt.cimt_24months-cimt.cimt_enrolment,
    vlSuppr_6months=ifelse(lims.HIV_Abbott_Viral_Load_0pt2ml_P_6months>150,1,0),
    vlSuppr_12months=ifelse(lims.HIV_Abbott_Viral_Load_0pt2ml_P_12months>150,1,0)
  )

# compare sub-/clinical event rates - not significant but 0 events in the small group of non-suppresseds
fisher.test(table(rhiccaDatWide$vlSuppr_6months,rhiccaDatWide$numClinEventsConfirmedPerParticipantCVDwithCIMT))

## 
##  Fisher's Exact Test for Count Data
## 
## data:  table(rhiccaDatWide$vlSuppr_6months, rhiccaDatWide$numClinEventsConfirmedPerParticipantCVDwithCIMT)
## p-value = 0.3779
## alternative hypothesis: true odds ratio is not equal to 1
## 95 percent confidence interval:
##  0.000000 2.985607
## sample estimates:
## odds ratio 
##          0

fisher.test(table(rhiccaDatWide$vlSuppr_12months,rhiccaDatWide$numClinEventsConfirmedPerParticipantCVDwithCIMT))

## 
##  Fisher's Exact Test for Count Data
## 
## data:  table(rhiccaDatWide$vlSuppr_12months, rhiccaDatWide$numClinEventsConfirmedPerParticipantCVDwithCIMT)
## p-value = 0.6092
## alternative hypothesis: true odds ratio is not equal to 1
## 95 percent confidence interval:
##  0.000000 4.180156
## sample estimates:
## odds ratio 
##          0

# compare CIMT at 24 months distributions - they are different!
wilcox.test(rhiccaDatWide$cimt.cimt_24months[rhiccaDatWide$vlSuppr_6months==0],rhiccaDatWide$cimt.cimt_24months[rhiccaDatWide$vlSuppr_6months==1])

## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  rhiccaDatWide$cimt.cimt_24months[rhiccaDatWide$vlSuppr_6months == 0] and rhiccaDatWide$cimt.cimt_24months[rhiccaDatWide$vlSuppr_6months == 1]
## W = 13564, p-value = 0.09548
## alternative hypothesis: true location shift is not equal to 0

wilcox.test(rhiccaDatWide$cimt.cimt_24months[rhiccaDatWide$vlSuppr_12months==0],rhiccaDatWide$cimt.cimt_24months[rhiccaDatWide$vlSuppr_12months==1])

## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  rhiccaDatWide$cimt.cimt_24months[rhiccaDatWide$vlSuppr_12months == 0] and rhiccaDatWide$cimt.cimt_24months[rhiccaDatWide$vlSuppr_12months == 1]
## W = 12766, p-value = 0.00626
## alternative hypothesis: true location shift is not equal to 0

# compare CIMT at enrolment distributions - they were already different at enrolment (at least for those still not suppressed at 12 months)
wilcox.test(rhiccaDatWide$cimt.cimt_enrolment[rhiccaDatWide$vlSuppr_6months==0],rhiccaDatWide$cimt.cimt_enrolment[rhiccaDatWide$vlSuppr_6months==1])

## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  rhiccaDatWide$cimt.cimt_enrolment[rhiccaDatWide$vlSuppr_6months == 0] and rhiccaDatWide$cimt.cimt_enrolment[rhiccaDatWide$vlSuppr_6months == 1]
## W = 21203, p-value = 0.08225
## alternative hypothesis: true location shift is not equal to 0

wilcox.test(rhiccaDatWide$cimt.cimt_enrolment[rhiccaDatWide$vlSuppr_12months==0],rhiccaDatWide$cimt.cimt_enrolment[rhiccaDatWide$vlSuppr_12months==1])

## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  rhiccaDatWide$cimt.cimt_enrolment[rhiccaDatWide$vlSuppr_12months == 0] and rhiccaDatWide$cimt.cimt_enrolment[rhiccaDatWide$vlSuppr_12months == 1]
## W = 16887, p-value = 0.01285
## alternative hypothesis: true location shift is not equal to 0

# compare CIMT at 24 months distributions - no difference in change-from-baseline values
wilcox.test(rhiccaDatWide$cimtChange[rhiccaDatWide$vlSuppr_6months==0],rhiccaDatWide$cimtChange[rhiccaDatWide$vlSuppr_6months==1])

## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  rhiccaDatWide$cimtChange[rhiccaDatWide$vlSuppr_6months == 0] and rhiccaDatWide$cimtChange[rhiccaDatWide$vlSuppr_6months == 1]
## W = 11815, p-value = 0.9338
## alternative hypothesis: true location shift is not equal to 0

wilcox.test(rhiccaDatWide$cimtChange[rhiccaDatWide$vlSuppr_12months==0],rhiccaDatWide$cimtChange[rhiccaDatWide$vlSuppr_12months==1])

## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  rhiccaDatWide$cimtChange[rhiccaDatWide$vlSuppr_12months == 0] and rhiccaDatWide$cimtChange[rhiccaDatWide$vlSuppr_12months == 1]
## W = 10180, p-value = 0.9504
## alternative hypothesis: true location shift is not equal to 0

## confirming this -- not signifant when accounting for baseline CIMT in a regression model (both with 24 months CIMT and change-from-baseline CIMT as response)
summary(glm(cimt.cimt_24months~cimt.cimt_enrolment+vlSuppr_6months,data=rhiccaDatWide))

## 
## Call:
## glm(formula = cimt.cimt_24months ~ cimt.cimt_enrolment + vlSuppr_6months, 
##     data = rhiccaDatWide)
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)         -0.024979   0.028268  -0.884    0.377    
## cimt.cimt_enrolment  1.115494   0.054756  20.372   <2e-16 ***
## vlSuppr_6months     -0.003974   0.013330  -0.298    0.766    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.008364014)
## 
##     Null deviance: 7.6789  on 501  degrees of freedom
## Residual deviance: 4.1736  on 499  degrees of freedom
##   (394 observations deleted due to missingness)
## AIC: -971.87
## 
## Number of Fisher Scoring iterations: 2

summary(glm(cimtChange~cimt.cimt_enrolment+vlSuppr_6months,data=rhiccaDatWide))

## 
## Call:
## glm(formula = cimtChange ~ cimt.cimt_enrolment + vlSuppr_6months, 
##     data = rhiccaDatWide)
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)  
## (Intercept)         -0.024979   0.028268  -0.884   0.3773  
## cimt.cimt_enrolment  0.115494   0.054756   2.109   0.0354 *
## vlSuppr_6months     -0.003974   0.013330  -0.298   0.7658  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.008364014)
## 
##     Null deviance: 4.2128  on 501  degrees of freedom
## Residual deviance: 4.1736  on 499  degrees of freedom
##   (394 observations deleted due to missingness)
## AIC: -971.87
## 
## Number of Fisher Scoring iterations: 2

summary(glm(cimt.cimt_24months~cimt.cimt_enrolment+vlSuppr_12months,data=rhiccaDatWide))

## 
## Call:
## glm(formula = cimt.cimt_24months ~ cimt.cimt_enrolment + vlSuppr_12months, 
##     data = rhiccaDatWide)
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)         -0.026199   0.028649  -0.914    0.361    
## cimt.cimt_enrolment  1.117985   0.055424  20.172   <2e-16 ***
## vlSuppr_12months    -0.001892   0.014465  -0.131    0.896    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.008402589)
## 
##     Null deviance: 7.6689  on 499  degrees of freedom
## Residual deviance: 4.1761  on 497  degrees of freedom
##   (396 observations deleted due to missingness)
## AIC: -965.68
## 
## Number of Fisher Scoring iterations: 2

summary(glm(cimtChange~cimt.cimt_enrolment+vlSuppr_12months,data=rhiccaDatWide))

## 
## Call:
## glm(formula = cimtChange ~ cimt.cimt_enrolment + vlSuppr_12months, 
##     data = rhiccaDatWide)
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)  
## (Intercept)         -0.026199   0.028649  -0.914   0.3609  
## cimt.cimt_enrolment  0.117985   0.055424   2.129   0.0338 *
## vlSuppr_12months    -0.001892   0.014465  -0.131   0.8960  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.008402589)
## 
##     Null deviance: 4.2157  on 499  degrees of freedom
## Residual deviance: 4.1761  on 497  degrees of freedom
##   (396 observations deleted due to missingness)
## AIC: -965.68
## 
## Number of Fisher Scoring iterations: 2

RHICCA: analyses for paper 1

Marc Henrion

22 August 2024

1 Introduction

2 Abbreviations

3 Study objectives

4 Changes from the Statistical Analysis Plan

5 Set-up

6 Descriptive analyses

6.1 Data overview

6.1.1 Data completeness & loss-to-follow-up

6.1.2 Loss to follow-up

6.1.2.1 Study flow diagram

6.1.3 Participant baseline data and demographics

6.1.3.1 CMV and VZV categories

6.1.4 Multiple imputation

7 Baseline population have a low cardiovascular risk profile.

7.0.1 Modifiable CVD risk factors and inflammatory and viral markers over time

8 Risk of asymptomatic CVD: surrogate markers CIMT and PWV

8.1 Carotid intima-media thickness (CIMT)

8.1.1 Descriptive summaries

8.1.2 Regression models

8.2 Pulse Wave Velocity (PWV)

8.2.1 Descriptive summaries

8.2.2 Regression models

8.3 Combining CIMT and PWV on a single table

8.4 Repeating this for all participants

8.5 Repeating this while including ART treatment

9 Summary of clinical and sub-clinical events

10 Incidence rate of clinical and subclinical events are ~2-folds higher in PLWH.

10.0.0.1 Extra figures for manuscript

10.0.0.2 Viral control

10.0.1 Hypothetical vaccine trial

10.1 All participants

10.2 PLWH

10.2.1 Without CD4

10.2.1.1 Sensitivity analysis

10.2.2 Table 3 for manuscript

10.2.2.1 CIMT

10.2.2.2 CIMT (with ART)

10.2.2.3 PWV

10.3 Time-to-event models

10.4 Multi-panel figure CIMT, PWV and CVD events

11 Incidence rate of systemic events (mostly hypertension and hypercholesterolaemia) was prevalent in both groups.

12 Mediation analysis

12.1 Y = asymptomatic CVD (continuous CIMT), M = ICAM-1, X = participant type (HIV)

12.2 Y = asymptomatic CVD (continuous CIMT), M = ICAM-1, CD163, IL-6, X = VZV status (PLWH only)

12.3 Y = clinical and subclinical (CIMT) CVD events, M = ICAM-1, CD163, IL-6, X = VZV status (PLWH only)

13 Additional things for discussion